Tactiq works by reading the captions your video platform already shows, so no bot joins and no audio is recorded. For Google Meet that is a Manifest V3 extension: switch captions on, watch the caption element, store finished lines with speaker and time, export Markdown and send the text to a model for a summary. An agent builds it in one sitting. The catch is upkeep: the extension depends on Meet's page markup, which changes without notice, and each extra platform (Zoom, Teams) is another set of selectors to keep alive.
Before you buildCheck the free plan first: Tactiq Free keeps 10 transcripts a month with sharing and export, which may be all you need.
Build me a Chrome extension that transcribes my Google Meet calls, to replace Tactiq. Requirements: - Manifest V3 extension in TypeScript, bundled with esbuild. `npm run build` writes ./dist/, which I load unpacked from chrome://extensions. No UI framework: plain DOM for the side panel and options page. - Capture: a content script on meet.google.com turns on Meet's captions, watches the caption container with a MutationObserver and stores finished lines as (time, speaker name as Meet shows it, text). Meet rewrites a line while someone speaks, so commit it after 2 seconds without change or when the speaker changes. - Put every DOM selector in selectors.ts with a comment each, and add a "Self-test" button on the options page that reports which selectors still match, because Meet's markup changes without notice. - Storage: IndexedDB through `idb`: meetings (title from the page, start, end, participants) and lines. Nothing leaves the browser profile. The service worker closes a meeting when its tab closes. - Side panel (chrome.sidePanel): live transcript, a Highlight button on any line, a list of past meetings and full-text search over them. - Export: Markdown with title, date, participants and highlights on top and the transcript below, saved through chrome.downloads, plus a plain .txt. - Summary on demand: a button sends the transcript to an OpenAI-compatible endpoint (base URL, model and key set on the options page and kept in chrome.storage.local; Ollama works) and saves a summary, decisions and action items with owners next to the meeting. - No accounts, no telemetry, local-first. The only network call is that endpoint, and only when I press Summarise. - Out of scope: Zoom and Teams, joining as a bot, recording audio, speaker identification beyond Meet's names, sync. - README: why Meet's captions must stay on, loading the unpacked extension, and a reminder to tell participants.
Opens your agent with the prompt filled in, you press enter. Or copy it and paste it into an empty folder.
The first prompt builds the core. Run these one at a time in the same folder to grow it into your own Tactiq.
Extend the Meet transcriber in this folder. Requirements: - Add a content script for the Zoom web client (app.zoom.us/wc) that follows the same capture flow: captions on, observe, commit finished lines. - Keep Zoom's selectors in their own file with the same self-test button, and show which platform each meeting came from. - Meetings from either platform share one list and one search. - Update the README, including that the desktop Zoom app is not covered.
Add automatic saving to the Meet transcriber in this folder. Requirements: - On the options page, let me choose a folder through the File System Access API and keep the handle in IndexedDB. - When a meeting ends, write YYYY-MM-DD-title.md there with the same Markdown layout as the export. - If the permission has lapsed, show a one-click "Grant folder access again" button in the side panel and queue the files meanwhile. - Update the README.
Add a question box to the side panel of the Meet transcriber in this folder. Requirements: - Find the 12 best-matching lines across all meetings with a simple BM25 ranking over the stored text, in the browser. - Send only those lines with their meeting titles and dates to the configured endpoint and ask for an answer that cites each source meeting. - Show the cited lines under the answer and never send meetings that were not matched. - Update the README.
One install covers Meet, Zoom and Teams without a bot in the call, and someone else keeps it working when the platforms change their pages. Teams add sharing, shared AI answers across meetings and integrations that push notes into the tools they already use.
Estimate · The extension runs in your browser at no cost. A summary through an API costs cents per meeting, or nothing with Ollama. The build is token usage on your agent plan.
| Plan | Monthly | Annual, per month | What you get |
|---|---|---|---|
| Free | $0 | $0 | 10 transcripts and 5 AI credits per month, AI-generated insights, in-meeting AI, share and export transcripts. |
| Pro | $12 | $8 | Unlimited transcripts, 10 AI credits per month, pause automatic transcription, option to disable the transcript notification. |
| Team | $20 | $16.67 | 1 to 20 users, unlimited AI credits, priority support, auto-share transcripts with teams. |
| Business | $40 | $29.17 | 20 to 200 users, advanced data retention, SAML SSO, customer success manager, Tactiq MCP and Claude connector (beta). |
| Enterprise | custom | — | 200+ users, a curated data processing agreement, customizable SAML SSO and in-meeting notification, dedicated account manager. |
Free gives 10 transcripts and 5 AI credits a month for one user, with sharing, export and in-meeting AI.
Yes. An AI coding agent such as Claude Code, Codex or Cursor can build a usable personal version in one session with the prompt on this page, running on your own machine.
About $12 a month on the Pro plan per seat, $144 a year (checked 2026-10-19).
Zoom and Microsoft Teams transcription, which need their own selectors; Fixes when a platform changes its captions markup: that job becomes yours; Workflows, integrations with Drive, Notion and similar tools, and team spaces; Sharing links, speaker statistics, cross-meeting AI search and the daily digest; Transcripts in other languages beyond what Meet's own captions support. If any of those matter to you, keep paying.
Yes: Meetily (open source), Anarlog (open source), Vexa (open source).