Changelog

A new AI Focus Groups template runs the same probing, concept-testing focus-group conversation with every participant individually, at scale, with a live demo you can try right on the page.

• Template pages now show what's actually inside each one before you use it: the fields it captures, example questions, and how it adapts and routes, with the most common questions answered inline.
• Security details across template and product pages now reflect our full posture, including SOC 2 Type II and ISO 27001:2022, with a link to the Trust Center.

Template page titles no longer repeat the brand name, so browser tabs and search results read cleanly.

Behind-the-scenes improvements to keep your workspace running smoothly.

• Analysis sessions can now chart your conversation data as line, pie, stacked bar, scatter, and heatmap visualizations, not just bar charts and tables, so trends, proportions, and outliers are easy to spot at a glance.
• You can now create multiple MCP keys and scope each one to specific workspaces, replacing the single all-access token. Connect Claude, Claude Code, or Cursor to only the workspaces a key allows, and revoke keys individually without disrupting your other connections.

When you make a copy of an agent, you can now keep its original owner instead of taking ownership yourself. You're added to the copy as a member, so you keep access either way.

The primary button in your research sidebar now reads New agent instead of New conversation, making it clearer that you're spinning up a new AI agent.

Behind-the-scenes improvements to keep your workspace running smoothly.

• Have a specific Perspective question? Every docs page now has an Ask Perspective button. Click it to ask the Perspective assistant directly and get a guided answer drawn from our docs, so you don't have to read through them to find what you need.
• The Perspective SDK now supports a "push" mode for the slider embed. Instead of overlaying the page, the slider takes its own column and lets the rest of the page stay fully interactive.

Perspective's marketing site, docs, and blog are now properly discoverable to AI assistants and crawlers. Every page advertises pointers to our machine-readable resources, and blog and docs pages expose a clean text version of their content so agents do not have to parse HTML.

Signing up no longer breaks if the welcome interview captures a name with unusual characters. Your name is cleaned up automatically and you land in your workspace instead of getting bounced back to sign-in.

Your workspace view is now a clear list of all your agents (Interviewer, Concierge, Evaluator, and Advocate), each showing its type and whether it's a draft, ready, or active, with a new filter to narrow the list by agent type.

• Two new models for your text interviews: pick GPT-5.5 or Gemini 3.5 Flash from your workspace's text interviewer model setting.
• Text interviews running on Gemini reply faster, with thinking depth tuned to keep reasoning quality high.
• The Concierge designer now pulls brand assets and a website overview as two clear cards when you give it your URL. Brand assets refresh on demand, and the website overview falls back to a live homepage scan when web search comes up empty.

The Slack channel picker recovers from connection errors. After reconnecting Slack and clicking Retry, your channels load and the picker becomes interactive, instead of staying stuck on the error.

Settings has been redesigned with a dedicated sidebar, clearly grouped sections, a one-click way back to your research, and a consistent layout on every page.

• Slack channels deep in the list are reachable again. Channel search now works across your entire workspace, even on workspaces with thousands of channels, so you don't have to scroll or click "Load more" to find one near the bottom.
• Designing a conversation no longer triggers stray "connection timeout" popups, and the hidden Try-it-out preview won't silently start a voice call in the background.
• Text conversations no longer flash a blank frame between the welcome screen and the conversation itself. Your greeting and first reply stay visible the whole way through.

• The embed setup guide now walks you through the tabbed embed panel step by step, so the instructions match what you see when you set up an embed.
• New step-by-step guidance on refining your outline from past conversations, with copyable examples for pointing the design agent at a transcript.

Stop manually writing outline updates. The conversational agent designer can now read your past conversations and propose outline changes based on what actually happened. Ask it to look at a previous conversation, and it reads the full transcript and writes the updates for you.

• Docs answers in the designer chat now start streaming within a second, instead of waiting 10 to 20 seconds for the answer to appear.
• The designer now spends more reasoning effort on the hard parts of your question and moves quickly through the simple parts.
• Embed setup is easier to navigate, with tabs for preview, code, coding agent, and customization. The new coding agent tab ships a copyable prompt that points Cursor or Claude Code at the Perspective MCP server for your real embed snippet.
• You can now send files and screenshots in chat without typing a message, and the dictate and send buttons no longer compete for the same slot.

The knowledge section of generated outlines stays focused. It used to pick up behavioral instructions, form fields, or participant criteria; now it only carries the product, business, or policy facts you provided.

Perspective now supports Gemini 3.5 Flash. Pick it from the model menu in the chat input. Stronger reasoning, a 1M-token context window, and faster output mean the design AI can think harder over longer sessions without making you wait.

Perspective conversations (Interviewer, Concierge, Advocate, Evaluator) sound less like a chatbot in voice and text. They stop narrating what they're doing internally, stop parroting your words back, and stop re-asking for things you've already told them.

• Phone calls flow better at topic transitions. The AI now jumps in with a quick acknowledgment while it puts together a full reply, which should reduce long silences.
• The AI now handles phone calls and text chats differently. On phone, it skips questions that don't translate well to voice, like asking you to spell your email, and uses your phone number from the call itself.
• Speakerphone calls are smoother. The AI no longer mistakes its own voice for you trying to interrupt, so it stops cutting itself off mid-sentence.

• Your dashboard now shows what actually happened on each phone call. You can tell whether the person hung up after wrapping up, went to voicemail, didn't answer, or hit a busy signal, instead of every call looking the same.
• Teammates whose email contained capital letters could be locked out of perspectives they should have access to. That's fixed. Owners of older perspectives can also reliably access their own work.

Your whole team can now edit a perspective together, not just the owner. Teammates and workspace members can change settings, manage automations and webhooks, chat with the design AI, invite more teammates, and mark interviews as test or rejected. Deleting a perspective and changing ownership are still owner-only.

When a workspace is out of credits, teammates who aren't workspace members now see "ask a workspace member to add more" instead of an "Add Credits" button they couldn't use.

The Form Data tab on an interview now shows "No form data extracted from this interview." when no fields were filled in, instead of going blank.

Interview summaries now skip empty form fields, whether you're reading them in-app or in a member summary email. The focus stays on the values participants actually provided.

Shared interview and changelog links now show their preview card again on Slack, X, LinkedIn, and other social platforms.

Buying credits works again on first navigation into the app. The Select Payment Method screen now loads the card form right away instead of showing a blank dialog with a dead Continue button.

• Your Concierge, Interviewer, Advocate, or Evaluator no longer paraphrases the welcome message you wrote in its first message. Participants see your welcome and then get fresh dialogue that builds on it instead of echoing it. On Slack, WhatsApp, and phone conversations, where the welcome message isn't displayed separately, your agent includes the welcome's intent in its first message.
• Behind-the-scenes improvements to voice conversation startup.

Behind-the-scenes security updates to keep your workspace safe.

Spend less time waiting after you hit the Talk button. Voice conversations now connect about 15x faster than before.

Perspective conversations embedded on your site now load about 3x faster, so visitors spend less time staring at a loading state and more time actually talking.

Behind-the-scenes security updates to keep your workspace safe.

• Curious what a Perspective conversation sounds like over the phone? Give our demo line a ring at +1 (559) 785-2453 and have a live voice conversation with our concierge. You'll be able to give your own conversations a phone number soon.
• Speak instead of type. The dictation mic now shows up on every chat input across Perspective, so you can dictate your prompts anywhere you'd normally type one.

Voice conversations sound more natural. The AI no longer slips in awkward filler before it answers or acts, and it stays in the same language for the whole conversation.

Slack messages sent by your automations now display their full content in the conversation's activity: text, attached blocks, button labels, and emoji all show correctly. Before, the activity often said "No message content".

Voice conversations now use a smarter realtime model with a context window four times larger for longer, more coherent back-and-forth. Live transcripts run on a new streaming-tuned model for sharper, lower-latency captions.

Voice interviews on iOS no longer get stuck in a non-scrollable state on first load, and the animated background now stays pinned to the screen instead of drifting with scroll.

Behind-the-scenes security updates to keep your workspace safe.

When the design chat agent picks a research methodology, the response now wraps cleanly on narrow screens instead of breaking into squeezed columns.

• Long research questions are faster to say than to type, and you can ramble through them without compressing your thought into something easier to type. The new mic in the design chat input (or press Cmd/Ctrl+Shift+D) transcribes you in real time and splices the words in at your cursor, so a long prompt takes the time it takes to speak it.
• Ask "how does X work in Perspective?" inside the design chat instead of opening another tab to dig through the docs. The assistant now searches the public docs and FAQ when you ask product questions and cites its sources inline, so you can verify or jump to the full page in one click. Docs citations link to the source page; FAQ citations expand into a popover with the question and answer.

Refreshed the docs across collect, design, and settings so guides match the current product.

Improved request reliability so the app no longer occasionally stalls on the first request after a period of low traffic.

Behind-the-scenes security updates to keep your workspace safe.

Grok 4.3, xAI's latest reasoning and vision model, is now available across the AI chat and analysis flows, and as the default model for data extraction. A faster low-reasoning variant is available for text interview deployments where response latency matters more than depth. Existing sessions on the previous Grok 4 generation transparently route to Grok 4.3.

The research analyst chat now keeps time-based horizontal bar charts in chronological order. When the AI charts data across weeks, months, or sequential stages, the bars stay in order instead of being re-sorted by value.

Behind-the-scenes security updates to keep your workspace safe.

Perspective MCP responses now include each perspective's internal name alongside its public title, so AI agents can identify perspectives by the label you recognize in the app.

Monospace text across the app now renders in DM Mono, giving code blocks, keyboard hints, and similar elements a consistent typeface that pairs with the existing DM Sans body font.

• HubSpot form automations now correctly link each submission back to the participant's anonymous browsing session, restoring page-view and source attribution on the resulting HubSpot contact.
• Perspective MCP outlines that include navigation steps with templated redirect or submit URLs now load reliably in strict clients like Claude Desktop, instead of failing with a tool execution error.

• Redesigned AI tool calls to look more consistent. Generating outline, classifying research, generating a greeting, evaluating an outline, saving an automation, picking a Slack channel: every action now flows through one consistent card with crisp loading, success, error, and stopped states, expandable details, and inline actions that keep the next step one click away.
• Results and Interviews dashboards load noticeably faster, especially on busy workspaces.
• Workspace credit status is now visible to everyone with access to the workspace, not just admins. The plan label and credit chip sit side by side in the workspace switcher, with clear "Checking credits" and "Credits unavailable" fallbacks so the chip never silently disappears mid-load.
• Refreshed the webhooks guide and automation recipes to match the product as it actually works today: the real event names, the real payload shape, the custom auth header model, idempotency keys, retry behavior, and the trust-score filter gotcha that silently excludes abandoned interviews.

Entry redirects on the Try Concierge page now preserve every query parameter, even when they contain encoded characters.

Sign-in popups in embedded conversations now load reliably on Safari.

• Perspective MCP tools now ship with complete behavior hints (read-only, open world, destructive), so MCP clients can show clearer confirmations before running each tool.
• Behind-the-scenes improvements to keep your workspace running smoothly.

Behind-the-scenes improvements to keep your workspace running smoothly.

• A polished new reading experience for blog posts. Long-form articles now feature an "In this article" table of contents that anchors directly to each section, one or two context-relevant product suggestions woven into the reading, and a curated list of related reads in the same category at the end. Section headings click-to-anchor cleanly under the top navigation.
• The blog home is now Field notes on conversational AI. The index reads like an editorial publication, with the same category bar and post grid you already use.

Shared interview link previews on social platforms now reliably show the host's avatar.

Preview interviews now unlock automatically when you add credits to your workspace. Previously, previews you started before purchasing a credit pack could stay locked even though credits were available.

Perspective is now agent-ready. AI assistants and search crawlers can discover what's available on Perspective's site (including the Perspective MCP server), and every blog, docs, and changelog entry is now also available as a plain markdown version that AI tools can fetch and cite directly.

• You can now add a description to each workspace in workspace settings (or when creating a new one). AI assistants connected through Perspective MCP use what you write to pick the right workspace for the task, so be specific about what each workspace is for ("growth research" vs. "support feedback") instead of relying on the name alone.
• The Slack channel picker on the invite page now loads instantly even on large workspaces.
• Thirty-six new posts on the blog covering AI customer engagement, research, insurance, intake and lead qualification, onboarding, customer success and churn, voice of customer, product discovery, and surveys.

• Slack integrations are now more reliable: invites and automations consistently use the active workspace connection, including right after reconnecting a Slack workspace.
• The trusted-companies logo strip on the homepage and the comparison pages now lines up with the rest of the page content on wide screens.

• Each use-case page now reads like a guide to solving a specific business problem with Perspective: leading with the value, walking through real features alongside photos and integrations, and pointing to a clear next step.
• You can now add Perspective MCP to Claude.ai on the web as a custom connector. The setup guide has the deep link and the exact fields to fill in.
• Twenty new posts on the blog:
  • Customer success and churn:
    • AI for Customer Success Is Stuck on Dashboards. The Real Unlock Is Conversations.
    • Customer Success Automation in 2026: The 4-Layer Stack Every CS Org Needs
    • How to Identify At-Risk Customers Before They Churn (A 2026 Playbook)
    • How to Reduce Customer Churn in 2026: A Modern SaaS Playbook
    • Churn Prevention Software in 2026: A Buyer's Guide to the Right Stack
    • Customer Churn Prediction with AI: Why Models Alone Aren't Enough in 2026
  • AI conversations:
    • AI Feedback Collection: From Static Surveys to Conversations That Actually Tell You Something
    • AI vs Surveys: Why Conversations Win for Real Customer Research
    • AI Qualitative Research: A Practical Guide for Modern Research Teams
    • Replacing Forms with AI Chat: When, Why, and How to Make the Switch
  • AI-native customer engagement:
    • What "AI-Native Customer Engagement" Actually Means (And Why Most Vendors Get It Wrong)
    • AI Customer Engagement Tools in 2026: A Buyer's Roundup by Use Case
  • Intelligent intake and vertical playbooks:
    • AI-Native Onboarding Software: What to Look For in 2026
    • Best AI Tools for Insurance Brokers in 2026: A Practical Roundup by Workflow
    • AI for Insurance Agencies in 2026: From Lead Capture to Renewals
    • AI for Real Estate Agents in 2026: A Practical Playbook for Top Producers
    • AI Client Intake for Law Firms: How to Replace PDF Intake Forms with AI Conversations
  • Product discovery and UX research:
    • UX Research at Scale: How AI Interviews Break the Researcher Bottleneck
    • User Interview Software in 2026: A Comparison Guide for Modern Research Teams
    • AI Product Feedback Tools in 2026: A Buyer's Guide for Product Teams

Working with Perspective MCP from Claude, ChatGPT, Cursor, and other clients is more predictable now: conversation totals match between tools, statuses come back in the same casing everywhere, invalid date filters surface a clear error instead of being silently ignored, and looking up a missing invite returns one predictable message. AI clients also pick the right tool for the right agent more reliably, since tool descriptions now reflect the four agent types: Interviewer, Concierge, Evaluator, and Advocate.

You can now connect Claude, ChatGPT, Cursor, VS Code, and other MCP clients to your Perspective workspace through standard OAuth, with no more pasting access tokens by hand. When an app asks for access, you grant or decline on a clean consent screen with the requested permissions spelled out in plain language. A new Connected Apps section in your workspace settings shows every authorized app and lets you disconnect any of them in one click. See the MCP setup guide.

Polished how Perspective MCP exposes automations to AI agents. Managing automation rules now uses dedicated list, create, update, delete, and test tools instead of a single catch-all, and tool titles and descriptions are sharper across the surface, so agents calling them behave more predictably.

• The context window indicator in research and analysis chat no longer flickers off the moment you send a new message. It now keeps the last known size visible until the next reply lands.
• Smoothed several inconsistencies across Perspective MCP tools: searching perspectives by title finds matches even when a custom display name is set, conversation totals match between the get and stats tools, the integrations tool recognizes every connected provider including Gmail, and listing conversations now signals whether more results exist alongside the cursor.

• Welcome chat now surfaces the interactive quick-reply panel and a Talk button. Participants can answer single choice, multi choice, and scale questions right in the welcome flow, or jump straight to a voice conversation without typing.
• When you create a new agent by uploading a brief, form, survey, or playbook, the system now reads the file alongside your text prompt to pick the right agent type for you.

• Interactive questions now support more options. Single and multi choice questions can show up to 10 choices, and scale questions can span up to 20 points, so the AI can present richer panels when the question calls for it.
• Copy-paste embed snippets now include a Perspective AI attribution comment above the script tag, so AI crawlers that can't execute JavaScript still see the attribution.

• HubSpot per-interview submissions are reaching HubSpot again after a brief regression, and the activity timeline now accurately shows failed submissions instead of masking them as successful.
• Brand logos in the interview header are now sized to be legible on both mobile and desktop. Configurations with a single uploaded logo keep a subtle backdrop for contrast; multi-logo and themed configurations drop the invisible frame so the logo has room to breathe.

Perspective MCP is now more reliable and handles long, multi-step perspective creation sessions without hitting tool timeouts in Claude Desktop and other MCP clients.

• The mobile navigation menu now adapts to your sign-in state. Signed-in users go straight to their dashboard, and signed-out users land in the sign-up flow.
• Polished the interactive question panel in text conversations. Hover and keyboard focus now share one selection ring so mouse and keyboard navigation stay aligned, scale endpoints read as labeled captions beneath the row, and the panel border highlights when focus moves inside.

Internal links in blog and docs posts now open in the same tab; external links continue to open in a new tab.

The homepage, blog, docs, and marketing pages are now blazingly fast. Pages load instantly from cache and stay smooth as you scroll and interact. You'll feel the difference the moment you land.

Behind-the-scenes improvements to keep your workspace running smoothly.

Interactive question panels: AI conversations can now present participants with single choice, multi choice, and scale/NPS questions as native UI panels instead of plain text. Includes keyboard navigation and "something else" free-text option. Messaging channels (Slack, WhatsApp) fall back to quick-reply chips.

Behind-the-scenes improvements to keep your workspace running smoothly.

New blog post: "The Glasswing Principle: Why Your Customer Feedback Tools Have the Same Blind Spot": a fresh take on the limits of today's AI feedback tools.

• Restored the Use Cases page to show only the original 25 curated items instead of all 181 templates.
• Lighter breadcrumbs for snappier page loads.
• Faster post-login redirects with a leaner workspace lookup.

• Long-running analysis chats no longer break. Sessions are now durable: you can close your browser while analysis is running, come back to finished results, and resume an in-progress analysis from any device.
• Faster navigation across results, design, and conversation pages.
• SDK 1.7.0: auto-injected attribution signals so AI crawlers (GPTBot, ClaudeBot, PerplexityBot) can discover Perspective AI on your site. Enabled by default with a one-line opt-out.

• Faster results page navigation.
• Performance tuning across conversations, automations, and analysis sessions.
• Better citation chip contrast in light mode; preview conversations are now included in redirect count checks.
• Now serving llms.txt to follow the emerging convention for AI-discoverable content.

Routine security and dependency updates to keep your data safe.

• Template gallery expanded to 181 pages with strong search visibility and a clean category system.
• New conversation-based template cards with dynamic previews and avatars.
• Templates now appear in the global navigation (header, mobile, footer).
• Slack profile context is now included in AI prompts for richer participant context.

• Redesigned templates gallery with sidebar categories and grouped sections.
• Template detail pages and embed previews use the live SDK widget for a richer experience.
• Simplified citation chips in analysis sessions.

• Cleaner whitespace in participant nudge emails.
• Faster template search and general UI polish across cards, sidebar, and embeds.
• Only one embed is active at a time when the preview modal is open.

Analysis now works across hundreds of interviews

• The assistant runs qualitative theme discovery and quantitative pattern analysis in the same conversation.
• Search: Semantic search across all interview transcripts. Ask natural language questions and get answers with citations linking back to specific participants.
• Data analysis: Quantitative queries across your structured output fields (counts, distributions, trends, cross-tabulations).
• Deep exploration: Full-text search across all transcripts to surface patterns, specific quotes, or emerging themes that structured fields miss.

• Citations now show participant initials and first names as clickable chips.
• The assistant refreshes its knowledge of your interviews on every turn, staying current on participant counts, date ranges, and metadata.

• Trusted per-invite participant context: invite links can now carry verified context (URL parameters, invite metadata, Slack profile) that's stored securely and merged into the participant's session.
• Increased URL parameter value limit from 20 to 50 characters.
• The invite tool now validates context and prompts you when something is missing.

• Legacy metadata entries and empty values are now handled safely.
• Empty metadata wrappers are no longer emitted when there are no valid entries.

Routine security and dependency updates to keep your data safe.

• Faster interview research page loads thanks to leaner data fetching.
• SDK 1.6.1: fixed avatar padding on the float launcher bubble.

Automation test sends (digest, per-conversation, invite) now correctly deliver to the test recipient instead of the configured recipients.

Routine security and dependency updates to keep your data safe.

New text conversation models: GPT-5.4 (mini, nano) and Claude Haiku 4.5 for faster, cheaper AI conversations.

Routine security and dependency updates to keep your data safe.

Analysis session visibility: admins can hide sessions from customers with an admin-only / everyone toggle. Admin-created sessions default to hidden, with instant UI feedback and direct-URL access enforcement.

• Three-tier conversation completion handling: conversations are now classified as complete, partial, or abandoned, with the right depth of AI processing for each tier. Slack notifications get colored sidebars and emails adapt to status.
• Completion source tracking: manual, participant, and no-activity completions each get distinct messaging. Background processing now prevents duplicate runs.

Routine security and dependency updates to keep your data safe.

• Voice instructions: researchers can customize voice accent and speaking style under Settings > Voice & Tone or through design chat. Applied in voice mode only.
• Persisted embed settings with a redesigned accordion UI: appearance, launcher, and auto-trigger options now save to your workspace and can be updated through design chat with an explicit publish step.
• SDK 1.6.0: applies your saved embed settings automatically (taking precedence over inline embed code) and replaces the loading spinner with a polished skeleton UI.

Routine security and dependency updates to keep your data safe.

Voice conversations no longer end prematurely when a participant asks for time. The AI now treats pauses as natural moments to wait, not signals to end the call.

Routine dependency updates.

• More reliable markdown link rendering across the marketing site.
• Aligned footer styling with top navigation for visual consistency.
• Logo marquee now scales more responsively across screen sizes.

• Restored links that were incorrectly blocked by overly strict link validation.
• Fixed image aspect ratio warnings in the logo marquee.
• Restored the missing footer top border.
• Silenced a smooth-scroll console warning during page transitions.

• Blog and docs now feature syntax-highlighted code blocks, copy buttons, and improved heading anchors aligned with the table of contents.
• React SDK documentation added to the Embed API reference.
• 17 new long-form blog posts covering Intelligent Intake, AI customer research, AI conversations at scale, and product discovery and UX research.

• Consistent agent styling: unified colors and variants across dropdowns, chips, and buttons, including per-agent colors in suggestion chips.
• Stronger search coverage with internal linking across the template gallery and cross-linking between blog posts.
• Better accessibility: aria labels and focus states for heading anchors.

• Agent colors now render correctly in dropdown menus.
• Fixed mobile overflow issues in docs pages and enabled horizontal scrolling for code blocks on mobile.
• Resolved duplicate anchor and override issues in heading links.
• Cleaned up blog content artifacts: unicode escaping, truncated descriptions, and incorrect template mappings.

SDK 1.5.1: float button now expands to fullscreen on mobile.

• Voice conversations no longer time out on long sessions, thanks to a leaner conversation prompt.
• Conversations can now resume when a participant switches back to text after ending voice mode.
• Fixed an iOS welcome chat scroll overlap on small screens.

• Rewrote the Typeform alternatives blog post for stronger search visibility, with refreshed metadata, internal links, and updated pricing and positioning copy.
• Perspective AI now ranks #1 in the Typeform alternatives list with a unified comparison view.

New global error page preserves URL query parameters on reload, so tracking and redirect context aren't lost after recovery.

• Docs reorganized around what you're trying to do: Overview, Design, Collect, Analyze, Settings, with a new "How Perspective Works" intro, agent-based framing, and core guides covering all four agent types.
• AI-native doc actions: every docs page has a "Copy page" dropdown to export as Markdown or open it in ChatGPT, Claude, or Gemini with a prefilled prompt.
• Expanded developer docs: API overview, webhooks, automation recipes, and Embed API reference.

• Redesigned docs homepage with a quickstart CTA and a two-column layout; sidebar now uses shorter navigation labels.
• Refined CTA button styling: consistent outline design with improved spacing, hover states, and sizing.
• Stronger search visibility: structured data for articles, how-tos, and videos, plus accurate sitemap priorities and last-modified dates.

• Docs correctness: fixed broken links, typos, API values, and webhook auth docs; updated Outline docs to match the actual UI.
• AI copy/export: prevented duplicate page titles and corrected title handling.
• Date and timezone fixes for search-engine and sitemap output.

• Outline evaluator: new tool reviews research outlines with full context and streams its reasoning and evaluation as it works.
• Selected tags now show on conversation cards and the details page, backed by a centralized tag vocabulary and reusable badges.

• Claude Sonnet 4.6 context window increased from 200k to 1M tokens, supporting much larger inputs.
• Faster, smoother voice-to-text handoff: mode switching is now instant with no blocking revalidation between modes.

Voice dark theme is now scoped to the voice experience, preventing premature theme changes elsewhere.

• HubSpot automation submissions no longer silently fail: the portal is now resolved automatically from your connected account, integration errors are detected properly, and bot-protection warnings now surface to users.
• Partial-conversation polish: stale processing records no longer crash, cleanup runs in the background for lower latency, and the progress bar no longer shows 100% for partial conversations.

Routine security and dependency updates to keep your data safe.

AI conversations now use correct grammatical gender for self-reference in gendered languages like Hebrew, with voice mode picking gender automatically from the selected voice.

AI streaming now reports errors consistently so failures surface instead of silently dropping.

Switched to a more stable streaming transport to prevent crashes from malformed responses.

Voice conversations now use dark mode from the very first screen, including welcome and sign-in, for a consistent, polished feel across previews and embedded dialogs.

Contact and newsletter forms now use the standard embed widget so theming and host handling stay consistent.

• iOS conversations no longer auto-scroll unexpectedly when the address bar or viewport shifts; user scroll position is preserved.
• Custom slugs with hyphens (e.g. /contact-us) now load correctly instead of failing to authorize.

Claim flow now supports multiple workspaces via an allowlist instead of a single hardcoded workspace.

Blog and docs now render from local content for faster page loads and reliable static generation.

Claim flow now shows a clear error state when no workspaces are available instead of hitting a dead end.

• Idle interviews now resolve automatically into Complete or Partial after a configurable timeout per research, so abandoned conversations stop hanging in progress.
• Participants can resume or restart abandoned and ended interviews across web, Slack, and WhatsApp with a clear Continue or Start Over choice.
• New per-research setting to remove Perspective branding across the interview UI, emails, embed sign-in, and previews.

• Refined messaging, timing, and researcher insights for partial interviews (previously called "abandoned") across email and Slack.
• AI tool error states now use a calmer neutral style instead of alarming red.

Branded email previews now reflect the remove-branding setting immediately instead of showing stale content.

• New participant invite tool generates magic-link invites and can email them directly.
• Float embeds now support launcher customization: pick an icon (or branded avatar), size, and shape, with live preview and updated embed snippets.
• Centralized FAQ system with structured data across homepage, pricing, compare, and agent pages for better search visibility.

• Redesigned embed Visual Options panel for clearer configuration.
• Embed config now exposes a branded avatar URL for launcher icons.

• Float embed preview bubble is now fully visible thanks to a taller preview frame.
• FAQ links no longer show a stray arrow icon, and CTA tracking now uses referral parameters.

A new claim link lets users claim a pre-built research perspective and copy it into their workspace through the standard sign-in flow.

• The claim flow now restricts cloning to workspaces you have access to.
• The interview footer no longer mismatches the surrounding theme.
• Pasted or uploaded attachments now stay visible even when the input is empty.
• Long attachments no longer cause horizontal page scroll in the interview view.
• Social link preview images now render correctly for image URLs without a file extension.

Routine security and dependency updates to keep your data safe.

New scrolling logo strip on the homepage, compare pages, and concierge trial page: 26 trusted company logos in a slow continuous marquee, with dark mode and reduced-motion support.

• Voice assistant no longer responds to its own audio output during conversations.
• Reduced transcription errors in noisy environments by tuning the transcription model and handling empty transcripts more gracefully.
• Fixed a small visual jump when the logo marquee loops.

Voice connection timeout extended from 10 to 30 seconds so slower networks don't drop calls during setup.

Routine security and dependency updates to keep your data safe.

• Voice calls stay connected through quiet moments — keep-alive pings now happen every 30 seconds instead of every 2.5 minutes to prevent idle timeouts on the upstream voice provider.
• Final voice events are no longer lost during call shutdown.

Redesigned interview status pages: polished cards replace the old emoji-based error screens, and invalid URLs now redirect to a dedicated marketing page instead of a dead end.

The top nav, sidebar, and footer now show the full set of roles and industries from the central use cases directory instead of a hardcoded subset.

"My perspectives" filter now correctly returns research for users with multiple emails, and deleted items disappear from the list immediately.

• New URL input on the comparison pages with validation and a responsive layout.
• Submitting a URL on a compare page now flows directly into the concierge experience, preserving attribution context through signup.
• Added a Compare link to the footer for better discoverability.

• Compare pages now match the design system: shared typography, theme colors instead of one-off values, and a simpler hero with headline, subline, and URL input.
• Refined comparison tables and competitor messaging across all compare pages, with stronger positioning of AI conversations vs forms.

• Interview chat no longer auto-scrolls to assistant messages before they have visible text, eliminating jumpy scroll behavior.
• The sticky footer now has an opaque background so message text doesn't bleed through underneath.
• Signed-in users who land on the sign-in or signup page without a redirect target now go to the research dashboard instead of the homepage.

Routine security and dependency updates to keep your data safe.

• Send participant email automations directly from your connected Gmail account, with built-in connection, validation, and permission setup.
• New sign-in banner above the chat input for authenticated conversations, translated across 18 languages.
• "Sent via Perspective" inline attribution appears when the email footer is hidden in branded Gmail mode.

• Sign-in-first welcome flow now requires authentication before collecting user info, with refreshed calls-to-action and translations.
• Participant email templates now render list fields cleanly as checklists, bullets, or comma-separated values based on context.

• The welcome flow now clears the sign-in prompt after authentication completes.
• Restored reply-to addresses on participant emails delivered by Perspective.
• Fixed missing accent marks in Spanish, Portuguese, Polish, and Turkish translations.

Routine security and dependency updates to keep your data safe.

• Docs now support deeper category nesting, with breadcrumbs, sidebar, and prev/next navigation that traverse the full tree.
• Docs categories without their own page now show a list of their child pages automatically.
• The marketing homepage now shows for everyone instead of redirecting signed-in users to the research dashboard.
• Simplified the mobile menu by removing the non-interactive Channels and Automations sections and giving "Get Started" a more prominent button style.

• Fixed invisible UI elements across marketing pages — badges, gradients, workflow diagrams, feature cards, and story-player effects — caused by bad theme color values.
• Fixed step indicators and connector lines on use-case detail pages so they render and stay aligned in their column.

Concierge email buttons no longer contain a malformed URL segment, so links now route correctly across all three concierge email types: created, no form found, and error.

• Redesigned homepage with a new narrative emphasizing the shift from forms to AI conversations, plus refreshed Problem, Solution, Benefits, Use Cases, and Automation sections.
• New Studies and Tutorials entries in the top navigation.
• New visualization elements for product demos: leaderboards, tag clouds, insight bubbles, and insight cards.

• Docs sidebar now uses a collapsible category layout with icons, faster preloading, and smarter highlighting of the current page.
• Standardized marketing calls-to-action and FAQ sections across landing pages for visual consistency.
• Faster page rendering thanks to parallel data fetching and removal of duplicate blog API calls.

• Fixed unreadable button text in dark mode and several interaction bugs in the docs sidebar.
• Fixed a brief flicker on some pages and a layout bug in the calendar component.

• Native Slack interview thinking and streaming, so participants see responses appear in real time.
• Channel @mention support: interviews now work in channels with the same active-conversation handling as direct messages, plus a branded call-to-action when no conversation is active.
• Slack file-share replies with text are now accepted (the attachments themselves are ignored).

• Slack mentions and links are preserved through automation and interview replies, including the final streamed message.
• Restarting a Slack interview now reuses the same interview but starts a fresh direct-message thread, cleaning up the old one.

Routine security and dependency updates to keep your data safe.

Embedded anonymous interviews now restore session state correctly when the embed reloads, so participants skip the welcome screen even when browsers block third-party cookies.

Routine security and dependency updates to keep your data safe.

• GPT-5.4 model support Added OpenAI's GPT-5.4 with a much larger context window, plus reasoning and vision. Now the recommended model in the model selector.
• Context usage indicator in chat A circular progress indicator in the chat input shows how much of the model's context window your conversation is using, with a tooltip for details.
• Voice interview controls Added mute/unmute and end-conversation buttons, a new "Left Early" status for participants who hang up, and clearer visual states for voice input.
• Stronger email verification Sign-in code throttling is now per-email and more reliable, with clearer feedback when codes are rate-limited and more secure code generation.

Preview interview sign-in When an existing session conflicts with preview interview sign-in, you now get a warning before the session is cleared.

Voice reconnect stability Voice connections no longer get stuck during startup or reconnects, with bounded retries, clearer error messages, and reliable recovery from connection failures.

• Interviewer now asks "anything else?" before the summary confirmation, so participants get a chance to add anything they missed.
• Tightened the closing flow so the interview only ends after all closing steps are complete, and the final message is always a statement rather than a question participants can't answer.

Prevented interviews from ending prematurely. Previously the closing message could contain a question participants had no way to respond to.

Voice interviews now wait for the assistant's closing message to finish playing before completing. Previously the connection could drop mid-goodbye, cutting off the final words to participants.

• Sign-in required interviews in embeds Authenticated interviews now work inside embedded iframes even when browsers block third-party cookies, using a sign-in popup with a new-tab fallback if popups are blocked.
• Auto-captured participant emails When participants share their email during a conversation, it's now captured automatically so thank-you automations work without extra setup.
• Compare hub page A new index page linking all competitor comparison pages, with a hero, stats, and category-grouped logo cards.

Concierge landing page polish: cleaner hero state, better typing indicator, and consistent text sizing in chat messages.

• Quick reply chips are now readable in light mode.
• Email button links no longer include a doubled domain.

Routine security and dependency updates to keep your data safe.

• Scheduled invite automations Send recurring invitations to participate in research conversations via email or Slack — perfect for daily standups, weekly check-ins, pulse surveys, sprint retros, and ongoing feedback loops.
• AI-composed invite messages Each invite is generated with the share URL and tailored copy, plus an interactive Slack channel picker in the assistant chat.
• Auto multi-conversation mode Invite automations automatically allow multiple conversations so each invitee can participate.

• Test mode for participant email automations now redirects to your override recipient instead of silently failing.
• The research assistant now offers to test automations in the first setup step, matching the later steps.

• Comparison pages New pages comparing Perspective AI to Typeform, Formstack, Qualtrics, SurveyMonkey, Google Forms, Jotform, and Medallia — each with a URL converter, feature table, FAQ, testimonials, and pricing details.
• Embedded newsletter signup Newsletter signup is now an embedded Perspective AI conversation instead of a static multi-step form.

• Outline generation restored Fixed errors that had been silently breaking outline generation for around a week. Outlines now generate reliably across all models.
• Improved dark mode contrast and accessibility on comparison page tables.

• No more duplicate interviews Background sign-in refreshes no longer restart interviews, so participants won't get duplicate sessions or restarted voice conversations.
• Outbound voice calls wait to greet The AI no longer speaks immediately on outbound calls — it waits for the recipient to speak first, which feels more natural for voice conversations.

No more corrupted chat messages Added a guard against rapid double-clicks and overlapping send-or-regenerate actions across all chat surfaces, preventing message corruption.

• Slack link previews restored Perspective AI URLs now show proper previews when shared in Slack — including blog posts, pricing, and the homepage — instead of appearing as plain links.
• Preview deployment URLs now generate link previews correctly too.

Routine security and dependency updates to keep your data safe.

• SEO overhaul across marketing pages Added social sharing metadata, structured data, and canonical URLs across the homepage, pricing, about, use cases, products, roles, and landing pages so they surface better in search and link previews.
• Cleaner page titles and descriptions Removed duplicate brand names, rewrote use-case titles to be keyword-targeted, and trimmed long titles and descriptions to fit search result limits.
• About page heading hierarchy Cleaned up the page heading structure for better readability and SEO.
• Renamed the blog from "Your AI Customer Compass" to "AI Customer Research Blog" for better discoverability.

Automation errors now display readable messages instead of empty objects.

Embeds now load your allowed channels and custom welcome message from the server, so they render channel-specific UI and show your welcome text out of the box.

Smarter auto-redirect Auto-redirects after interview completion now use the same URL handling as manual clicks for consistency, including search parameters and environment swaps.

• Welcome wizard timing The welcome screen now waits for participant info to be ready before processing, so fields no longer go missing on completion.
• Broader compatibility for participant data: common alternative field names for first name, last name, company, company size, and email consent are now recognized automatically.

• Richer blog link previews Blog posts now include structured metadata, social share images, and improved Twitter card details so they render beautifully when shared.
• Site overview for AI crawlers A new structured site overview, dynamically generated from published blog content, helps AI search tools index our content accurately.

• SOC 2 and ISO 27001 badges Added to the footer to surface our security certifications.
• Cleaner version list in the dropdown with proper scrolling and no viewport overflow.

• New landing pages for Intelligent Intake and CX Teams, each with hero, features, how-it-works, impact, use cases, FAQ, and CTA sections.
• Added interactive demos for modal previews, animated conversation playback, embed mockups, and routing, insights, and intent analysis visualizations.

Conversation redirects now consistently carry the conversation reference, so context is preserved whether the redirect happens early or at the end of the flow.

Fixed early navigation redirects dropping the conversation reference, which previously caused inconsistent landing experiences.

• Auto-trigger embed popups Popups can now fire automatically based on a delay or exit intent instead of requiring a button click, with frequency controls (always, per session, or per visitor) and code generation for Script, NPM, and React.
• Workspace invite roles When inviting teammates, you can now choose between Admin and Member with role capability descriptions inline.
• Faster voice model Upgraded to a new realtime voice model: better instruction following, +10% transcription accuracy, and improved tool calling at the same price for voice conversations.

• Reliable interview goodbyes Interviews now consistently deliver the closing message instead of skipping the farewell or showing raw tool output.
• Lower latency for AI responses Streaming responses arrive faster across interviews, research, and analysis, with reduced time-to-first-byte in multi-step tool flows.

Routine security and dependency updates to keep your data safe.

• Two-step conversation ending Conversations now end in a clean two-step sequence: the model marks the session complete, then delivers its goodbye to the participant, so sessions no longer get stuck open and farewells are reliably shown.
• Restored an additional model option for conversations now that the new ending flow handles its constraints.
• Right-to-left languages now flow correctly through entire action card headers.

Prevented sessions from getting stuck when models output internal commands as raw text. Participants now consistently see a proper goodbye at the end.

Routine security and dependency updates to keep your data safe.

• Per-agent prompt configuration Each agent now owns its own role, objectives, engagement rules, and methodologies. Cleaner separation makes agent behavior more consistent and easier to evolve.
• More reliable session endings Recovered cases where some models occasionally failed to end sessions properly, so conversations now close cleanly across more model variants.
• Clearer session lifecycle Models now reliably distinguish between wrapping up a conversation and ending the session, so sessions end when they should.

Embed integrations now correctly fire navigation callbacks instead of silently dropping them.

Standardized color usage across the app for cleaner visuals.

• Restored missing animations and theme accents that were lost in the recent design system update.
• Attachments now hide when voice mode is on and the input is empty.
• Mobile attachment button no longer renders when there isn't enough space.

• Run conversations over WhatsApp as a second messaging channel alongside Slack, with secure delivery, per-recipient status, post-completion feedback buttons, and clear error messages for participants.
• Added support for a new high-context reasoning model with a 1M token context window and configurable thinking depth, now the recommended choice.

• More responsive thinking state across AI chat and conversations, with the logo morphing into bouncing dots during reasoning.
• Refreshed styling across research, outline, admin, and marketing surfaces with a faster, more consistent design system.

• Fixed reasoning steps not rendering for the new high-context model.
• Fixed WhatsApp invite ownership and feedback confirmation consistency.

• Slack DM conversations Participants can now complete conversations directly in Slack DM threads, with branded DMs, channel broadcast announcements, and owner mentions in the invite.
• Slack channel option in invite wizard A flatter invite wizard with a new Slack channel option, permission checks, and a reconnect flow when access needs to be refreshed.
• Smarter Slack experience Permalinks open the Slack thread panel, hints mention participants, completed conversations get a thread reply, and every invite has a DM button.

• Automatic context trimming Long conversations now have history pruned automatically as they grow, preserving the last successful outline.
• Google One Tap is now hidden when a conversation doesn't require sign-in.

Conversations in Advocate now wrap up smoothly instead of cutting off abruptly.

Fixed a small bug in model selection.

• Global admins can now view, initiate, cancel, and send reminders for workspace ownership transfers.
• New call-to-action closing flow for Advocate conversations: a single CTA response instead of a summarize-and-confirm step.

• End-of-conversation text (thank-you, loading, CTA, redirect) now renders in the conversation language across 17 languages (English, Spanish, French, German, Hebrew, Arabic, Chinese, Japanese, Portuguese, Russian, Italian, Dutch, Korean, Polish, Turkish, Swedish, Hindi).
• Redirect destinations are now visually highlighted to make the target site clearer.
• Switched the default and recommended model to a faster, more capable option, with older model selections automatically mapped forward.

Routine security and dependency updates to keep your data safe.

• Outline settings UI refresh: renamed "Brief" tab to "Overview", tighter typography and spacing, scrollable sections, and a single unified "Add" button for empty sections.
• Outline versioning simplified: version dropdown now shows version numbers and "Latest" without per-version metadata clutter.
• Hide outline "Add" buttons while the chat is loading or streaming to prevent conflicting actions.

Research settings now correctly accept values that exactly meet the minimum length (e.g., 3 or 10 characters) for Title, Welcome Message, Brand Name, and other fields.

• New Advocate research type with dedicated prompts, methodology guidance, and a tailored experience for persuasion, objection handling, and message playbooks.
• Workspace-level model setting for text conversations, with owner and admin controls in Settings.
• Outlines now have a Knowledge field for structured factual context required during conversations, and can read attached PDFs and images.
• Outline change tracking shows who changed what and when, with version metadata and key differences in the editor.
• Refreshed model lineup with updated context windows and pricing, including transcription and text-to-speech.

• Updated conversation and research assistant prompts for advocate support and stricter knowledge boundaries.
• Research classifier now appears only when multiple research types are available for the chosen agent.

• Payment provider sync now uses the workspace owner's email instead of an admin's.
• Custom-domain surveys now fall back through a chain of detection methods, with AI as a last resort.
• Added support for multi-step survey navigation, including SurveyMonkey.
• Broken brand favicons are now cleared instead of retried indefinitely.

• Participant emails now show your brand logo, colors, and signature when brand mode is enabled, replacing the purple top bar with a clean, white branded design.
• Brand favicons now save reliably so they always render in admin previews and emails.

• Up to 85% faster first response after the opening message, thanks to smarter caching.
• Faster conversation startup: research data now loads in parallel with session lookups.
• Branded CTA buttons now pick light or dark text automatically for better contrast on any background color.

• Switching between voice and text no longer drops in-flight messages, with smoother transitions.
• Fixed stuck "Preparing your summary…" state when a reflection finishes without content.
• Smoother thank-you to CTA transition at the end of conversations.

Routine security and dependency updates to keep your data safe.

• New chat-first conversation experience: a voice-only mode with a centered microphone, and an inline "Talk" button that swaps with submit when typing begins.
• Localized placeholders and hints across 16 languages.

• Smoother end-of-conversation transitions, with a progressive loading indicator that eliminates layout shifts as reflections resolve.
• Tuned the welcome greeting animation timing for a snappier feel.

Voice-only actions like ending a conversation and extracting names now run reliably again.

Replaced the static contact page with an embedded Perspective interview that routes sales and support requests to the right destination and notifies Slack on completion.

Removed the dimming overlay from the mic and keyboard selection sheet for a cleaner experience.

Restored the external tool integration endpoint after a third-party update broke the connection.

Behind-the-scenes improvements to keep your workspace running smoothly.

• Stop button now cancels in-flight AI generation and tool runs, including intake form inspection and website analysis.
• Analysis sessions and the research assistant also support cancellation via the Stop button.

• Tool interfaces now clearly show streaming, interrupted, and error states.
• Activity tab now renders for conversations in any state.
• Smoother avatar uploads with built-in cropping.

• Switching between voice and text no longer drops assistant turns mid-conversation.
• Fixed a duplicate timestamp issue in avatar uploads.

• Renamed "Post conversation flow" to Automations and replaced collapsible sections with a tabbed layout (Per Conversation / Scheduled) for clearer configuration.
• Embedded conversations now forward all URL parameters from your page, not just UTM tags, so context flows through cleanly.

• Moved "Generate conversation summary" and "Generate trust assessment" toggles into Automations settings.
• The Add button tooltip is now context-aware based on the active tab.

Solution pages no longer load duplicate demo widgets: consolidated to a single responsive section.

Routine security and dependency updates to keep your data safe.

• New option to hide the fullscreen button in embedded conversations for a more contained experience.
• Added a "Fullscreen Button" toggle to embed appearance options so generated snippets reflect your choice.
• The Research Assistant can now answer questions about webhook setup, authentication, and event payloads.

Greeting generation now uses the configured display name for more accurate personalization.

• Complete redesign of the Solutions section with a Jobs-to-be-Done content model, filterable card grid, and shareable links.
• New Solutions detail pages with a two-column layout, sticky embedded demo conversations, and a 4-step "How it works" flow.
• New end-of-conversation action messages with completion CTAs, auto-redirects, countdowns, and thank-you states, including a reflection card that streams a participant-specific summary.
• Quick reply chips for text conversations, generated contextually per question.
• Calendly automation type added with its own icon and dedicated configuration.

• Solutions pages now load faster and rank better in search.
• Lighter-weight completion confetti for a snappier finish.

• Skipped automations no longer report as successful in test runs.
• More reliable streaming of reflections at the end of a conversation.
• Better text direction handling for multilingual conversations.

Improved voice conversation reliability so transcripts and audio always save.

More reliable voice connection startup: eliminated stale state issues that could cause the start guard to misfire.

Brand avatar now shows correctly in preview interviews when brand mode is enabled.

• New Agent mode for research creation: generates a complete research outline in one pass without clarifying questions.
• Research created via the API and external tool integrations now defaults to Agent mode; in-product creation remains interactive.

Social share image generation now handles a wider range of avatar formats by automatically converting them.

Routine security and dependency updates to keep your data safe.

• Brand mode for conversation identity: toggle between personal owner info and discovered brand identity, with automatic brand selection when website analysis finds brand data. Email templates respect the selected display identity.
• New Activity timeline tab on completed conversations shows post-conversation workflow steps (extraction, trust, automations, integrations) in real time.
• New animated processing indicator with logo and shimmer replaces the static "Processing…" text across conversation screens.
• The right agent type is now auto-detected and saved for older research projects.

• Workspace billing access restored: all members can view billing, credits, and subscription details again.
• Faster realtime voice responses thanks to quicker turn detection.
• Smoother avatar and logo uploads with built-in cropping and instant preview refresh.

• Fixed stuck loading states and stale data issues in brand mode.
• Streaming reliability improved with automatic retries and graceful backoff.

Routine security and dependency updates to keep your data safe.

• Fullscreen interview button in the header and floating controls: available for authenticated users and for anonymous users in embed mode.
• Tracking and context URL parameters are now preserved when opening an interview in fullscreen from an embed.
• Independent branding controls: custom logos and theme colors can now be enabled separately.

• Smarter logo and color extraction from analyzed websites, with better SVG handling, CAPTCHA detection, and favicon fallback.
• More concise responses from interview chats.

• Hebrew (right-to-left) titles now align correctly in URL previews.
• Anonymous sessions are now generated correctly when skipping welcome steps.
• Site favicon now loads reliably across the app.

• Automation failure tracking: in-app banners, inline error indicators in settings, and email alerts for persistent failures.
• Customizable call-to-action copy via a new actionText option on the research input (defaults to "Start research").
• Form inspector now extracts fields from custom web components, including Salesforce forms.

Errors are now classified as temporary versus permanent to drive smarter retry and notification behavior.

• Slack automations now survive channel renames by sending via channel ID instead of name.
• Underlying tool errors (e.g. "channel not found") now surface clearly instead of generic failure messages.
• More reliable file uploads: automatic client retries, fixed stuck spinners after failure, and multi-step embedded HubSpot forms now extract correctly.

Routine security and dependency updates to keep your data safe.

• New token-authenticated API endpoint to create conversations: POST /api/v1/perspective/create with token-based authentication.
• Auto-generated conversation titles and sensible defaults when creating via your AI assistant.

• Better first-message personalization: greeting generation now uses the user's message context.
• Shared default greeting and signature across the app and AI assistant tooling for consistent behavior.

Restored blog and docs content loading after a recent infrastructure change.

• Embedded conversation flow controls: Skip the welcome screen and start directly in voice or text, and immediately redirect on completion when a return URL is provided.
• Embed SDK migration: Embed now ships via published @perspective-ai/sdk and @perspective-ai/sdk-react packages with global CDN delivery for faster loads.
• UI theming: Custom theme support for chat input buttons and user message bubbles.

Simplified connection messages, tooltips, and prompts in the voice input for clarity.

Conversation header now always shows a logo, defaulting to the Perspective logo when no custom logo is set.

Multiple conversations per participant: Participants can complete a conversation more than once when enabled, supporting longitudinal studies, recurring feedback, and follow-ups. Each completed conversation creates a fresh attempt.

• Standardized toast notifications: Migrated to a consistent toast system across the app for clearer, more uniform feedback.
• Smoother credit-locked chat: Out-of-credits accounts now see a persistent warning banner above the input, with the input UI remaining visually consistent.

• Accurate trust labels in Slack: Trust labels in Slack messages now match the in-app score thresholds, preventing misleading labels for low scores.
• Conversation restart edge cases: Fixed priority and edge conditions around restarting conversations vs. starting new ones.
• No more horizontal scroll: Conversation link previews no longer scroll sideways on small screens.

Routine security and dependency updates to keep your data safe.

• All workspace members can manage billing: Buying credits, changing plans, and viewing invoices is no longer restricted to workspace owners.
• Clear "No Permissions" fallback screens appear in billing dialogs when access is missing.
• More reliable Stripe payment redirects and clearer error states in plan selection flows.

• Embedded conversation onboarding: New welcome flow for first-time users and a join flow for invited workspace members, both powered by AI-native onboarding with completion handling.
• Experience landing pages: New themed campaign pages that embed conversations directly (e.g., AI experience, product insights, TWIST).
• Embedded conversation widget: New widget with theme sync, lifecycle management, event callbacks, and brand customization support.
• Gemini 3 Flash (preview): Added across model selection.

Invited users are routed to the join flow and new users to welcome automatically.

Right-to-left languages now render correctly during streaming voice messages, spanning full width with proper alignment.

Routine security and dependency updates to keep your data safe.

• More reliable file uploads: Upload errors now return clear messages instead of failing silently, while keeping all existing safety checks.
• Cleaner conversation spacing: Adjusted message spacing so the thinking indicator no longer causes layout shifts.

• New participant-email automation sends AI-personalized follow-ups after conversation completion, with a preview dialog and reply-to support.
• Welcome screen now accepts a typed first response when voice is disabled, auto-submitting once the conversation starts.
• Embed SDK gains brand color customization and light/dark/system theme overrides, with branded loaders.

• Thinking indicator now stays visible for any assistant message with reasoning, while animated phrases play only during streaming.
• Email preview is more responsive on mobile, with default-template tooltips and cleaner setup.
• Automations keep running gracefully when individual steps fail or no conversations exist.

Fixed style bleed between light and dark modes in embed theming and corrected brand color formats.

• Fresh research images after upload: Avatars and logos now refresh immediately after replacement instead of showing the old cached version.
• Corrected a duplicate Polish translation in the thinking indicator.

• Sign-in redirects preserve destination: Unauthenticated users sent to sign-in now return to their original page after logging in.
• Updated Hebrew thinking indicator translation for clearer phrasing.

• Greeting generator in research chat: A dedicated tool generates greetings and titles automatically on the first outline, without overwriting later edits.
• Embed theming everywhere: Theme config now applies to embed root and container elements, not just the embedded frame, for consistent appearance across embed modes.

• Faster research list: Centralized redirect logic and streamed loading for quicker perceived performance.
• Conversation summaries now render with proper markdown in the updates view.

• Conversation completion redirects work again for anonymous users.
• Voice conversations on iOS no longer have scrolling blocked by the embed scroll workaround.
• Automations correctly return their actual ID after creation, eliminating "Automation not found" errors.

Routine security and dependency updates to keep your data safe.

Footer copyright year is now dynamic instead of hardcoded.

• Custom HubSpot tools work reliably: Tool registration is now per workspace, so custom HubSpot tools are consistently available.
• AI assistant responses include explicit URLs: Tools now return both share and direct URLs in their responses.
• Resolved immediate "Automation not found" errors right after creating an automation.

iOS embed scrolling fix: Embedded conversations on iOS now keep content in view instead of loading off-screen.

• Reliable conversation endings: The closing message and end-conversation action now happen together in a single response, preventing missed completions.
• Correct embed redirects: Parent page navigation only happens in true embed mode, preventing misdirected redirects elsewhere.
• Updated footer copyright year to 2026.

Routine security and dependency updates to keep your data safe.

• Embed brand colors: Customize embedded conversations with light and dark brand colors, plus forced light/dark/system theme via simple data attributes or programmatic config.
• Fullpage embed mode: New embed mode renders the conversation in the full viewport for dedicated conversation pages.
• Dynamic system prompts: Research chat prompts are generated per turn with versioned tracking for better adaptability.
• Multilingual thinking indicator: Conversation thinking indicator now supports multiple languages.
• Resource scope badges: Resource cards show "Shared" or "Private" badges for visibility at a glance.

Better brand logo extraction: Logos now load correctly from externally hosted sources (e.g., WordPress) and are auto-categorized as light or dark based on luminance.

• Post-conversation automations: Complete automation system with one-click integrations, tag-based triggers, durable workflows, and scheduled digests.
• Embed redirect handling: Embedded conversations can navigate the parent page with security validation and cross-origin support.
• Thinking indicators in text conversations: Extended thinking now appears in text conversations for more thoughtful AI responses.
• HubSpot custom forms: Direct integration for HubSpot form submission within automation flows.
• Email tool: Send emails directly from automation workflows.
• Webhook configuration: Set up webhooks dynamically with auth headers and built-in test capability.

• Cross-workspace search from your AI assistant: Search by name or title across all accessible workspaces, with paginated results and disambiguation metadata.
• Default workspace on create: Creating a conversation no longer requires specifying a workspace; it defaults to your primary one. Responses now include workspace info.
• Workspace context in responses: Create, respond, and update responses consistently include workspace ID and name.

• Clearer search behavior: List endpoints handle browse and search, with detailed lookup reserved for fetching by ID.
• Embed options panel reordered: preview and embed code first, appearance options at the bottom.

Correct access control in search: Visibility checks now use the actual visibility list, ensuring users only see researches they're permitted to.

• Embed appearance options: Control visual elements of embedded conversations directly from the embed box.
• New embed URL parameters: Hide the greeting or branding via URL parameters for tighter integration.
• Floating resources button: When the header or branding is hidden, a minimal floating button still gives access to resources.
• Hardened embed security: Tighter validation on cross-frame messages.

The floating resources button now only appears when there are resources to show.

More reliable research URL detection across the app.

• Fixed broken focus ring behavior on sign-in buttons and aligned spacing between sign-in and sign-up pages.
• Focus rings no longer appear immediately when dialogs open, only on keyboard navigation.

• Infographic generation from insight cards: Generate and view shareable infographics directly from cards.
• Default workspace lookup from your AI assistant: New tool to fetch the default workspace, plus a browser-friendly endpoint.
• Workflow run tracking: Improved navigation flow inside workflows with run tracking.

• Upgraded file storage: More reliable uploads with size limits and improved avatar handling.
• Google One Tap is now hidden in embed mode for a cleaner embedded experience.

• SVG uploads disabled for logos: Prevents potential script injection via uploaded SVG files.
• Custom logo dimensions and display issues resolved across the conversation UI.
• Conversation completion CTA stays active as expected.

• More reliable embedded conversations: Consistent anonymous IDs across actions and resolved embed rendering mismatches.
• Auto-redirect on completion now defaults on: New returnUrl query parameter as a fallback completion redirect.
• Better code blocks: Line numbers, theme-aware syntax highlighting, and improved copy experience.
• Improved right-to-left language support: Direction-aware padding, list styles, and outline rendering.

Mobile layout fixes: prevented badge wrapping and kept action menus within card bounds.

• Full conversation lifecycle from your AI assistant: New tools to create, update, respond, and set actions on conversations end-to-end.
• Conversation retrieval: List, fetch, and batch-fetch conversations with filters and stats breakdowns.
• Stats with average duration: Conversation stats now include average duration in seconds.
• Embedding support tools: New tools to fetch preview links and embed options for embedding flows.
• Webhook testing: New capability to validate webhook setup.

AI assistant tokens scoped to user: Tokens migrated from workspace to user scope for a more secure permission model.

• Owner names now display correctly instead of showing the workspace name.
• Outline updates now respect user feedback instead of ignoring it.
• Custom slugs no longer get saved as the research ID on invite.

• Embed conversations directly into your own pages, with an updated sharing experience.
• New website analyzer tool plus expanded theme options in research settings.
• Choose agents and switch between voice and text per research project.

Smoother post-conversation research settings, now with the ability to trigger webhooks when conversations complete.

• New unified plan selection dialog replaces the separate buy and upgrade flows across the app.
• Rebuilt pricing page with tabs, dynamic starter prices, and adjusted Pro tiers.
• Pro subscriptions can now be started, updated, cancelled, and resumed from one place.

• Checkout now supports coupons, invoice previews, and a clearer success screen, with an enhanced Stripe payment method UI.
• Versioned credit policies, including zero-cost research and analysis, with smarter consumption logic.
• Workspace switcher now groups workspaces, prefetches prices, and shows plan badges.

• Sign in faster with Google One Tap and a new email verification code flow that replaces magic links.
• Send personalized interview invites with automatic reminders, plus a participant summary captured at completion.
• Pick from multiple OpenAI voices with sample playback, and preview interview resources in a new sidebar.
• Extract structured data from interviews at scale with a new AI-powered tool.
• Analysis answers now cite their sources with clickable links back to the original interview content.

• Redesigned sidebar navigation, refreshed empty states, and a simpler welcome wizard for new users.
• Better mobile layouts across the interview viewer, admin panel, and chat, plus stronger right-to-left language support.

Audio now plays reliably on iOS, transcription accuracy improved, and several scroll, dialog, and mobile layout glitches are resolved.

New research type system with clearer categories: Exploratory, Evaluative, Concept Testing, and Generative.

• Replaced auto-scroll with a manual scroll button so you stay in control of the conversation view.
• User messages now resize dynamically for easier reading, and the sidebar layout is simpler.
• Renamed "highlight" to "insight" across the app for consistent terminology.
• Public insight pages now show the analyst's avatar and name for proper attribution.
• Gift credits are consumed automatically when granted, removing an extra step.

Resolved a major bug in interview prompt handling, plus several mobile sidebar and scroll button issues.

• Sign in with LinkedIn for faster account creation.
• New trust assessment scores conversation credibility across multiple dimensions automatically when conversations complete.
• Mark conversations as test or rejected and filter by status, plus filter conversations by date range with fiscal quarter support.
• Public, shareable highlight pages with table of contents, hero images, and rich link previews.
• Edit highlights, titles, and descriptions inline, and use charts, tables, and a calculator directly in chat.

• Smarter research creation flow with an expert mode for advanced users.
• Voice conversations now save audio in MP4, auto-play next clips, and detect speech turns more accurately.
• Analysis model picker is grouped by family, and Claude Opus 4 is now available.

Fixed voice recording cutoffs, mobile scroll issues on results pages, trust assessment streams, and several workspace and outline glitches.

• Conversation progress bar now stays at least one-third full so progress always feels visible.
• Simulated conversations now lean toward positive sentiment and you can generate up to five per research project.

Resolved share copy button issues, fixed delete analysis session, and prevented overly long greeting messages.

Added timestamps to conversation transcripts and a visual indicator for natural pauses between messages.

• Simulate conversations with AI, complete with reaction tracking and summary emails.
• Marketing blog now supports tagging, categories, and cover images.

• Refreshed welcome prompt, multi-line template card descriptions, and clearer list headers.
• Smoother link sharing and improved anonymous interactions.

• Fixed scrolling in the highlight sidebar and the analyst assistant edit view, plus duplicate buttons in lists.
• Cover image and remote image rendering on post pages now works correctly.

• Anonymous access is now the default mode, so participants can jump in without signing up.
• Auto-generated welcome messages greet users at the start of every conversation.
• Share with teammates from the settings sidebar, plus easier public page sharing from list view.

• Major UI refresh for more consistent styling and better accessibility across buttons, inputs, dialogs, and dropdowns.
• Scroll mode is now the default in conversation chat, with smoother animations and a refreshed welcome screen.
• Renamed "Perspectives" to "Home" and simplified the homepage layout.
• Default credit allowance increased to 250 with an updated pricing page.

Resolved conversation assistant scroll positioning, mobile card layouts, confetti in scroll mode, and various hover and toggle button glitches.

• Toggle between "My Perspectives" and "All Perspectives" on the list page.
• Sort and filter your research list by status from new dropdowns.
• New workspace shared page for published conversations.
• Monthly credits allocation for workspaces.
• Share with teammates directly from the settings sidebar.

• Major speedup on list and search loading, tuned for snappier results.
• Refined outline generator experience and conversation preview, including hover tooltips and clearer cards.

• Redesigned Plan and Billing section with a cleaner credits counter and a manage button.
• Tier name now appears next to the Pro plan, and you can cancel Pro directly from the change plan dialog.
• Smoother mobile experience for conversation details and a simpler attachment upload indicator.
• Comprehensive email template redesign with stronger information hierarchy and a new gray-logo signature.
• Hebrew lists in the conversation guide now align correctly.

Fixed broken preview link in the research-ready email and ensured at least one credits package always appears.

• New lifecycle emails for plan downgrades, resumptions, low-credit alerts, purchase confirmations, and tier changes.
• Added Gemini 2.0 integration with an updated pricing model.

• Apply coupon codes during checkout.
• New in-app feedback dialog so you can share thoughts without leaving the product.
• Account lifecycle emails for signup, out-of-credits, purchase confirmations, plan upgrades, and tier changes.

• Stronger contact form validation, consistent request type design, and better dialog handling.
• Newsletter system now sends acknowledgment emails and internal notifications with cleaner wizard validation.
• Credits now track plan and expiration dates, and consumption logic prefers non-expired same-plan credits.
• Polished plan dialogs, centered credits counter, centered sign-in popup, and a more responsive about page.

Better error messaging for missing user accounts, fixed form layout glitches, and corrected upgrade subscription handling for edge cases.

Cleaner visual proportions on the about page and a more readable founder story.

Fixed the disabled tier selection dropdown and resolved stale content glitches in the blog.

Blog posts now support custom slugs, plus an option to bypass stale content when publishing.

• New tiered flow for Pro plan subscriptions.
• Contact form with submission handling, persistence, and notifications.
• Workspace auto-generation in the welcome wizard with unique naming, defaulting to your full name.
• Trusted company logos now appear on the signup page.

• Sidebar now reflects account type, surfaces quick actions, and shows unconsumed credits in a progress bar.
• Refreshed homepage copy, new about page design, and updated customer success and sales hero images.
• Featured use cases added to the website footer with a refreshed social proof section.

• Better handling of credit errors in analysis session creation and the credits counter now only counts open balances for the current plan.
• Restored chat input suggestion behavior when the input has a value, and tightened access control on the research route.

Homepage solutions section now showcases the first three featured templates.

Fixed a typo in the CTA section.

Updated the pricing page and removed an unused notifications dependency.

• Credit consumption and enforcement now apply to analysis sessions and research features.
• Locked input mode for analysis sessions with admin override.
• Restored the locked conversation behavior.

• Refreshed pricing page, solutions pages, and sales hero image, plus a solid-background website footer.
• Reordered the top menu, smaller trusted company logos, and a gradient added to the blog page.
• Suggestions moved into the chat input for quicker access.

Fixed home link display in header and footer, FAQ credit card typo, and ensured deleted conversations no longer count in perspectives list cards.

Routine security and dependency updates to keep your data safe.

Newsletter form with new user subscription support.

• Renamed "interview" to "conversation" and "messages" throughout the product, with consistent UI text and labels.
• Refined messaging in the sharing dialog and an enhanced single conversation view.

Restored simulator functionality.

• New announcements admin and a website banner so you can broadcast updates to visitors.
• Drag-and-drop sorting for templates makes organization easier.

• Updated pricing page layout with rounded card backgrounds, refreshed borders, and a layered hero image.
• Sticky navigation buttons, badge labels on solution cards, and overall navigation polish.
• New about page with initial structure and layout in place.

Fixed the header overlay issue with the solutions menu and a use case label typo.

• New workspace configuration step in the welcome wizard, with team size now optional in account settings.
• Email consent controls in your profile and welcome flow, plus an unsubscribe link in onboarding emails.
• Attachments support in research creation and new analysis sessions.
• Refreshed website with a new pricing page, centered solutions design, blog in the header menu, and gradient hero sections.

• Fade-in outline animation, redesigned highlight cards with repositioned tags, and template badges with dark mode support.
• Mobile menu now keeps the solutions list persistent, and the website footer has a cleaner layout.
• Research pages gained use case badges, a "Watch a Demo" button, and a responsive "Talk to Us" button.

Fixed website rendering issues and signup redirect problems.

• New workspace setup step in the welcome wizard for business users.
• Email consent controls in your profile and onboarding flow, plus an unsubscribe link in onboarding emails.
• Attach files when creating research or starting a new analysis session.
• Added Gemini 2.0 Flash as an experimental model.

• Outline templates now connect to the website solutions pages.
• Subtler outline fade-in animation and a cleaner AI model selector.

Profile settings now refresh session data automatically when the page loads.

• Read PDFs directly in chat with new PDF support in messages.
• Add attachments throughout the product, including research chat and analysis session chat.
• Auto-generated session titles, plus asset and highlight counts on session cards.

• Refreshed research outline view, AI model selector, and use-case icons in the new research page.
• Updated prompt system to allow questions, and removed the three-question limit from outlines.
• Shorter welcome CTA on mobile, default invite share message, and improved research and settings page titles.

Resolved errors that affected analytics settings.

Added unique page titles across documentation and blog for better search visibility.

• Attach files during conversations and view them in the conversation viewer.
• Paste files directly into chat with a refreshed input layout.

• Admin sidebar now matches the cleaner Research sidebar behavior.
• Better SEO with corrected canonical URLs and validated landing pages.

Resolved canonical link handling issues.

• New AI-generated landing pages with a unified view to manage them all.
• Customer success landing page variations now available.

• Refreshed landing page design and management views.
• Faster, more responsive conversation and preview experience.
• Outline version controls restored and improved.

• Experimental AI survey builder for creating forms.
• New analyst assistant sidebar with highlights and chat integration.
• Streaming charts and tables in analysis tools with version history.
• Redesigned outline creation experience and research settings.
• Automatic tags and descriptions generated for analysis tools.

• Outline now reliably delivered by the AI model.
• Welcome page redirects properly to the new creation flow.

• Preview emails directly in the admin interface.
• Email template management is better organized with icons for missing templates.

Major platform upgrades for improved performance and reliability.

Highlights "By Category" mode now displays correctly on mobile.

• Conversation preview now closes automatically after submitting feedback.
• New Day 7 email template with refined campaign linking.

Corrected typo in solution hero title.

• New math tool integrated into analysis chat with full calculator coverage.
• Edit highlights with an improved interface, including title and description editing in a sidebar.
• Highlight categories (tags) for better organization.
• Markdown content editing with a save experience for highlights.

• Highlight card width and table overflow issues resolved.
• Smoother category collapse transitions.

• Outline now lives in the sidebar with faster performance.
• AI tool generates outlines automatically.
• Anonymous users can set a name on the create page.

• Smoother outline streaming without flickering.
• Renamed "Engine" to "Model" throughout the app for clarity.

• Login flow no longer breaks when validating waitlist users.
• Selected AI model now carries over from the new analysis page.

New employee retention landing page.

Outline parse errors no longer break the list page.

• Upgraded to the latest Claude Sonnet model.
• Name validation now accepts unicode characters.
• Unauthenticated users land back on the welcome flow after signing in.

Friendlier subject line on the welcome email.

Embedded videos on the website no longer open at a tiny size.

The "new" button on the list page now opens the perspective creation flow directly.

• New welcome flow with a Day 1, Day 4, and Day 7 onboarding email sequence.
• Create perspectives directly from templates, with admin tools to manage them.
• Anonymous welcome and research design now available.

• Streamlined navigation with "New" as a primary menu item.
• Better mobile experience on the new page and home page scroll.
• Clipboard copy now plays nicely with Google Docs.

Deleting a conversation from the results page redirects properly.

• Fixed a broken support link.
• Deleting a conversation from the results page no longer redirects to the wrong place.

Upgraded AI models for better performance and stability.

• Add context URLs to perspectives with a new resources tool, including versioning and a header dropdown.
• Sign in with Google or other providers now opens in a popup for a smoother flow.
• Terms and conditions added to participant signup with a legal disclaimer on login.

• Smoother handling when sign-in popups are cancelled or hit errors.
• Resolved signup tracking and sign-up flow issues.
• Edit button in the list page now works correctly.

Improved primary and background colors in dark mode.

First-time users can now skip the initial setup wizard step.

• New welcome experience with refreshed visuals and an animated wizard.
• Profiles now support multiple job descriptions instead of one.
• Home page redesigned with a mobile-friendly hero section.

• Resolved conversation login and sign-in layout issues.
• Fixed assorted bugs in the welcome wizard.
• Various mobile look and feel improvements, including landing pages.

• Question outlines now handle malformed input gracefully.
• Resolved a bug in the perspectives list.

New perspectives list with a polished empty state.

Conversation summaries are now sharper and higher quality.

• Preview dialog when sharing links inside a conversation.
• Per-page video support, including an employee engagement video on the HR landing page.
• New testimonials section on the website.

Conversations now wrap up properly when they reach their end.

• Refreshed hero copy, FAQ content, and the "How it Works" section.
• New theme colors and updated partner logos (Berkeley, Stanford).

Launched the first landing page for employee experience teams, with three layout variants for testing.

• New "How It Works" section to explain the product end to end.
• Dedicated customer and employee experience use cases for more tailored analyses.
• Fullscreen video playback (note: autoplay isn't supported in fullscreen).

• Smoother scrolling to the active solution role button and to the results menu on mobile.
• Clearer empty states in results and a refined solutions content hierarchy.

• AI model labels now display in full on the analysis view.
• Highlights section is hidden when there are no highlights to show.
• Conversations can be marked complete with the right permissions.

• New marketing homepage and solutions page with responsive layouts and category scroll navigation.
• Redesigned results overview with tab-based filtering.
• Added Google Gemini 1.5 to handle longer conversations.
• Analytics now include AI-generated summaries for richer insights.

• Refreshed "Getting Started" experience and homepage layout.
• Invite emails now reach recipients even if they aren't registered yet.

• Fixed responsiveness on the banner and solutions hero.
• Fixed assorted typos and minor layout glitches across the marketing site.

• Workspace members: invite teammates, resend invites, and track last activity.
• Research now defaults to workspace visibility, with controls to update it.
• Buy credits and billing pages are now open to all users.
• Simulated conversations now consume credits, including empty simulations.

• Cleaner publishing header and a left menu with hover indication and links.
• Refined end-of-conversation experience with more prominent "done" and "tell us why" actions.

• First version of the blog with shared header and footer.
• New documentation site with "On This Page" navigation.
• Pro badge indicator for upgraded accounts.

• More reliable sign-in sessions.
• Refreshed branding and richer link previews when sharing Perspective URLs.

Better handling of payment errors during checkout.

• New credits purchase flow with simpler options and an inline payment form.
• Added an undo for changes to a workspace's unique URL.

• Renamed "Billing History" to "Invoices" for clarity.
• Reorganized the admin menu and refreshed the account view.

• Conversation URLs always show the latest content for participants.
• Hebrew text in link preview images now renders in the correct direction.

• Pre-fill participant name and email via conversation URL parameters.
• New anonymous mode for external conversations.

• Stronger authentication for external participants.
• Generated feedback image is now attached to the conversation summary email.

Fixed caching issues with feedback images.

• Generate a shareable feedback image for any conversation.
• Researchers can now override the default conversation language, and the AI can switch languages mid-conversation.

• Research list now sorts by most recent activity.
• Cleaner analysis prompt rendering and small icon refresh across the app.

Conversations now start in the participant's preferred browser language.

Emails now include a trackable homepage link.

• Fixed several bugs in email templates.
• Fixed the research list filter for legacy members and owners.

• Add a custom message when sharing conversations.
• Custom messages are now supported on shared highlights too.

• Skip option in conversations so participants can move past sections they don't want to answer.
• Add URLs to research context for richer source material.
• Soft delete for conversations, with the ability to recover them later.

• Tooltip showing participant names in the conversation summary.
• Sidebar menu toggle moved next to the logo for a more consistent layout.
• Smarter title generation.

• More resilient handling of bad outlines to prevent crashes.
• More accurate URL scraping when source pages contain partial lines.

• Share conversations via QR code.
• Completion emails now go to all Perspective members.

Removed autofocus from chat inputs on mobile for a better typing experience.

• Fixed a typo in the invite teammates flow.
• Fixed an endless redirect loop on the results page.

Behind-the-scenes improvements to keep your workspace running smoothly.