teams-transcript-extractor

A Claude Code / OpenCode skill for extracting complete transcripts from Teams meeting recordings hosted on SharePoint/OneDrive.

Microsoft Teams records meetings and stores them in SharePoint/OneDrive. The transcript is visible in the browser but:

• Download is often blocked (permission denied)
• Copy-paste is incomplete (virtual scrolling only copies the visible chunk)
• No export button in Teams or Stream

This skill solves all three problems automatically.

How it works

1. Opens the SharePoint Stream video page
2. Auto-detects if the player loaded (or if OAuth is needed)
3. Clicks the transcript button
4. Scrolls through the transcript panel in chunks (600px at a time)
5. Accumulates ALL text entries using Set deduplication (handles DOM recycling)
6. Returns the complete transcript as plain text

The core insight: SharePoint Stream uses virtual scrolling in its transcript panel — DOM nodes are recycled as you scroll. Simple querySelectorAll only captures the currently visible 20-30 items. This skill connects via Chrome DevTools Protocol (CDP) to run JavaScript that scrolls, extracts, and accumulatessimultaneously — getting every entry regardless of DOM recycling.

Installation

1. Install the skill

# Clone into your skills directory
mkdir -p ~/.claude/skills
git clone https://github.com/stanley6635/teams-transcript-extractor.git ~/.claude/skills/teams-transcript-extractor

For OpenCode: clone into your opencode skills directory.

2. Prerequisites

• Node.js 21+ — for native WebSocket support (zero npm dependencies)
• Chrome — with remote debugging enabled

3. Set up Chrome with remote debugging

Option A: Use bb-browser (recommended)

If you have bb-browser installed, it handles Chrome profile management and authentication automatically.

Option B: Launch Chrome manually

# macOS
open -a "Google Chrome" --args --remote-debugging-port=9222

# Linux
google-chrome --remote-debugging-port=9222

# Windows
"C:\Program Files\Google\Chrome\Application\chrome.exe" --remote-debugging-port=9222

4. Log in to Microsoft 365

Open your SharePoint/OneDrive in that Chrome window and log in. The session persists for future extractions.

Usage

From Claude Code / OpenCode

Just paste a Teams recording URL and ask:

"Extract the transcript from this meeting: https://hoyagw-my.sharepoint.com/..."

The skill auto-detects, auto-recovers from OAuth issues, and handles everything.

From command line

# Find the tab ID first
curl -s http://localhost:9222/json/list | python3 -c "import json,sys;[print(t['id'],t['title']) for t in json.load(sys.stdin)]"

# Extract transcript
node extract.js --port 9222 --tab-id <tab-id> --output transcript.txt

What it handles automatically

Problem	Solution
Stream player not loading (OAuth expired)	Detects unhealthy page → navigates to SharePoint home → triggers OAuth flow → returns to video
Virtual scrolling truncates copy-paste	CDP chunked scroll + Set dedup captures all entries
No download permission	Extracts via DOM instead of file download API
Encrypted transcript API response	Bypasses API entirely, reads rendered UI
Different screen sizes / zoom levels	Scrolls by pixel height, adapts to any viewport

Verified

This skill has been tested and works on:

Meeting	Duration	Language	Entries	Chars
HSO Pulse Live Q&A	~60 min	English	474	57,449
新入职同事产品培训安排	~87 min	Chinese	331	27,227

Both used the same DOM structure ([role="list"] + [role="listitem"]), confirming cross-region consistency.

Limitations

• SharesPoint Stream UI changes — If Microsoft changes the transcript panel's DOM structure, selectors may need updating. Tested working as of June 2026.
• Requires Chrome — Uses Chrome DevTools Protocol; won't work with Firefox/Safari.
• Login required — You must log into Microsoft 365 once per Chrome session. MFA is supported.
• Not for CI/CD — This is a desktop automation tool, not a headless API. It runs in your local Chrome with your login session.

How to contribute

Found a case where it doesn't work? Open an issue with:

1. The URL (if you can share) or a description of the page structure
2. The browser console output
3. Any error messages

DOM structure changes are the most likely breakage — if you notice the transcript panel looks different after a Microsoft update, check if the selectors in extract.js still match.

License

MIT