May changelog: we got into Y Combinator S26

What's new on screenpipe this week?

Really appreciate everyone's patience, support, and feedback. This will be a great opportunity to turn screenpipe into an amazing tool for everyone.

The richest context inside isn't in your wiki or CRM — it's already on everyone's screen. AI agents can't automate work they can't observe.

screenpipe is the local-first memory layer that fixes that.

Please update the app, it might be outdated

Or download it

⭐️ New features

Chat history page with full-text search, bulk actions, brand-aligned tabs, and a sidebar entry — find any past conversation in one shot
Chat source citations — every answer now footers the screen, audio, or memory it used. Verify before you trust
Smart connection suggestions on the chat dashboard — surfaces relevant integrations for what you're trying to do
Codex-style follow-up steering — queue messages while the agent is still working, reorder or cancel before they run
Better message edit UX — editing a previous turn no longer collapses the surrounding context
Chat picker in the collapsed sidebar — switch chats without expanding

Live transcript for auto-detected meetings (not just calendar-joined)
Drag one meeting onto another to merge them — preserves titles, attendees, notes from both
Collapsible transcript search + caret-follow in the note editor
Toggle to skip auto-appending typed text to the meeting note (Settings → Meetings)
Toggle to disable the live-note notification if it's noisy
Speaker search uses diarization turns — lands on the exact moment a person spoke, not the whole chunk
Stall notifications when the live transcript stops making progress

Resizable left sidebar with a hairline drag handle
Faster app launch — heavy chat / rewind renderers (code blocks, mermaid, message list) defer until needed
Snappier chat history — debounced + cached refresh, no more stutter on switch
Tighter chat composer + suggestions — preset picker, suggestion rows, source cards re-spaced
Microphone echo cancellation — real AEC for mic capture (room echo + speaker bleed gone) on both Windows and macOS

Filter + sort controls in the connections list
Google OAuth unverified-app walkthrough — clean path through verification
Browser session reuse on Windows + smarter extension popup (no need to re-login on the agent browser)

Per-parameter overrides for event-driven capture intervals (--capture-interval-vision-event, etc.)
screenpipe survey command — send structured feedback from the CLI
npm install telemetry on Windows so we can see install-time failures

Meeting transcript no longer lost when live STT drops mid-meeting — backfills from buffered audio
Google Calendar OAuth reconnects work; events deduped across multiple connected accounts
Speaker delete no longer fails silently when audio chunk files are missing
Chat no longer goes blank after "Summarize with AI" on meeting notes
Removed broken direct OpenAI Realtime option from the live-engine dropdown
Meeting-detected notification collapsed to a single action button (less noisy)
Workspace-swipe markdown stability in meeting notes
Audio writes use INSERT OR IGNORE to fix transcript replace conflicts

Tray flips to Error when recording is denied on permission-denied early return (was stuck "ok")
macOS prefers bundled ffmpeg over PATH to avoid version mismatch
Focus-monitor identity matching no longer mis-attributes windows after a display change
AI proxy strips refusals, retries on 400 sampling errors, drops orphan tool messages
Pro-cloud audio migration respects explicit engine choice + resets on logout
Tinfoil text + image redactors get the cloud token wired through