How to Set Up a Meeting Transcript MCP Server for Claude and Coding Agents (May 2026)
Learn how to set up a meeting transcript MCP server for Claude and coding agents in May 2026. Connect AI assistants to your meeting data without copy-paste.
Maintouch
You’re switching between your code editor, Slack, and meeting notes trying to remember what the team decided last week. Your agent could answer that question for you if it had access to the actual conversation. The meeting transcript MCP server lets you stop copy-pasting context. Connect it to Claude, Cursor, or Windsurf and your agent can query decisions, action items, and technical discussions from every meeting you’ve captured.
TLDR:
- MCP servers connect AI assistants to meeting transcripts without manual copy-paste
- You need Node.js 18+, API credentials, and an MCP-compatible client like Claude or Cursor
- Self-hosted servers require maintaining uptime and auth as APIs change
- Spinach AI connects to Claude, Cursor, ChatGPT, and VS Code via OAuth in minutes
- Spinach AI provides SOC2, GDPR, and HIPAA-compliant access to your last 100 meetings
Understanding the Model Context Protocol for Meeting Data
MCP (Model Context Protocol) is an open standard that lets AI assistants pull live data from external sources without manual copy-paste. Rather than switching tabs to find context, an MCP server acts as a structured bridge your AI agent can query directly.
For meeting data, this gap is real. Coding agents only know what they’re given. Without access to product planning calls, design reviews, or sprint retrospectives, they operate without the context that actually drove technical decisions.
An MCP server for automated meeting minutes closes that gap. When your agent needs to know why a certain architecture was chosen, it can query real conversation data instead of asking you to reconstruct it from memory.
Prerequisites for Setting Up a Meeting Transcript MCP Server
Before touching any config files, make sure you have the following ready.
- Node.js 18 or higher (run
node --versionto check) - npm or npx available in your terminal
- API credentials or an OAuth token from your meeting transcript source
- A meeting tool that exposes transcripts via API (Spinach makes your last 100 AI meeting notes available through its MCP server via OAuth)
- Basic comfort with JSON, since you’ll edit a config file to register the server
- Claude Desktop, Cursor, Windsurf, or another MCP-compatible client installed
If you’re missing Node.js, grab it from nodejs.org. Everything else follows from there.
Meeting Transcript MCP Servers: Spinach vs Self-Hosted and Competitor Options
Several approaches exist for connecting meeting transcripts to MCP clients, but they differ dramatically in setup complexity, maintenance burden, model compatibility, and transcript quality.
Spinach AI MCP Server (Recommended)
Spinach connects directly to Claude, Cursor, ChatGPT, and Windsurf through a single OAuth setup, giving your coding agents and AI assistants live access to meeting transcripts, decisions, and action items across your last 100 meetings. Setup takes minutes, with zero custom infrastructure or ongoing maintenance. Spinach provides SOC2, GDPR, and HIPAA-compliant access out of the box—compliance that self-hosted solutions require you to implement and audit yourself.
Self-Hosted Options (Meeting BaaS)
The open-source Meeting BaaS server gives you full control but requires you to maintain uptime, handle auth token rotation, update dependencies as APIs change, and keep your infrastructure compliant with security requirements. Every API update from your meeting provider becomes your maintenance burden. For teams without dedicated DevOps resources, this maintenance overhead outweighs the flexibility.
Other Meeting Tools
- Otter.ai and Fireflies: Both offer API access to transcripts, but neither provides a native MCP server. You’ll need to build and maintain custom middleware to bridge the gap—adding the same maintenance burden as self-hosting without the transcript quality and decision extraction that Spinach provides.
- Read.ai: Exposes limited transcript data via API with no native MCP support. Manual configuration required, and model compatibility remains inconsistent across Claude, Cursor, and ChatGPT.
Bottom line: Spinach eliminates infrastructure overhead while delivering superior transcript quality and broader model support. If you need production-ready context without DevOps burden, Spinach is the only MCP server built specifically for this use case.
Installing and Configuring Claude Desktop for MCP Integration
Claude Desktop reads MCP server configurations from a single JSON file. Find it at ~/Library/Application Support/Claude/claude_desktop_config.json on Mac, or %APPDATA%\Claude\claude_desktop_config.json on Windows. Create the file if it doesn’t exist yet, then add an mcpServers block:
{
"mcpServers": {
"meeting-transcripts": {
"command": "npx",
"args": ["your-mcp-server-package"],
"env": {
"API_KEY": "your-api-key-here"
}
}
}
}
Keep secrets in the env block, not in your args array, since args can surface in logs. After saving, restart Claude Desktop and check the tools icon in a new conversation. If your server doesn’t appear, review logs at ~/Library/Logs/Claude/ for error details.
Setting Up a Basic Meeting Transcript MCP Server (Self-Hosted Approach)
Self-hosting gives you full control over how transcript data is processed and served. The Meeting BaaS open-source MCP server offers a working starting point.
Clone the repo and install dependencies:
git clone https://github.com/Meeting-Baas/meeting-mcp
cd meeting-mcp
npm install
Add your credentials to a .env file:
MEETING_BAAS_API_KEY=your_api_key_here
Start the server locally with npm run start, then register it in your Claude Desktop config using "command": "node" and the full path to the entry file.
The tradeoff: you own uptime, auth, and maintenance as APIs change.
Connecting Your Meeting Transcript MCP Server to Claude Desktop and Claude Code
Once your MCP server is running, connecting it to Claude is straightforward. In Claude Desktop, open claude_desktop_config.json and add your server under mcpServers. For Claude Code, pass the server via the --mcp-server flag or configure it in your project’s .mcp.json file.
After restarting Claude, your meeting transcript tools appear automatically in the tool picker. You can then query transcripts, pull action items and create Jira tickets, or search past decisions directly inside your coding or chat workflow.
Integrating Meeting Transcript MCP Servers with Cursor and Windsurf
Cursor reads MCP config from .cursor/mcp.json in your project root, using the same mcpServers structure as Claude Desktop:
{
"mcpServers": {
"meeting-transcripts": {
"command": "npx",
"args": ["your-mcp-server-package"],
"env": {
"API_KEY": "your-api-key"
}
}
}
}
Windsurf uses ~/.codeium/windsurf/mcp_config.json with an identical schema. Drop your config there, restart, and your meeting transcript tools appear in the agent sidebar automatically.
Spinach’s MCP server connects to Claude, Cursor, Windsurf, and ChatGPT from a single OAuth connection. Configure credentials once and the same transcript context flows across every agent you work with.
Testing Your Meeting Transcript MCP Server Integration
With your server registered and Claude restarted, run a few quick checks before trusting the connection in a real workflow.
Start simple. Ask your agent:
- “What were the key decisions from my last meeting?”
- “Find any mentions of [feature name] across recent transcripts.”
- “What action items came out of this week’s sprint review?”
If the agent returns accurate data, you’re connected. If it responds with a generic “I don’t have access to that,” the server isn’t registering correctly. Double-check your config file for JSON syntax errors and confirm the API key is valid.
Run one more test: ask the agent to reason beyond simple retrieval. Try “Based on recent product discussions, what should I focus on in this PR?” That tells you whether context is flowing in a way the agent can actually use.
Common MCP Server Setup Issues and Solutions
Most setup failures trace back to a small set of root causes. Here’s a quick reference:
| Issue | Likely Cause | Fix |
|---|---|---|
| Tools missing in client | Server not registered or JSON syntax error | Validate config with a JSON linter, then restart |
| Authentication failure | Invalid or expired API key | Regenerate credentials and update the env block |
| Connection timeout | Server not running or wrong command path | Confirm npm run start succeeds locally first |
| Env variables not loading | Secrets placed in args instead of env |
Move all keys into the env block |
If issues persist, check client logs directly. Claude Desktop logs at ~/Library/Logs/Claude/; Cursor logs appear in the Output panel under MCP.
Advanced Configuration: Multi-Meeting Analysis and Cross-Meeting Intelligence
Once basic connectivity works, two capabilities unlock real depth.
With Spinach’s MCP server, your agent can reason across a full transcript history rather than just the most recent call. Track how a requirement evolved through multiple sprint reviews, or surface past decisions from Teams meetings that conflict with current code. The more meetings captured, the stronger the context your agent works with.
For recurring intelligence, Spinach’s Multi-Meeting Agents generate automated reports across configurable time windows (one day to one month), syncing with Glean and other knowledge bases, delivered via Slack or email. Every agent session starts with fresh, synthesized context without manual prep.
How Spinach AI’s MCP Server Provides Enterprise-Grade Meeting Context for Coding Agents
Spinach AI’s MCP server gives Claude, Cursor, and other coding agents direct access to your meeting history, decisions, and action items without any manual copy-paste. Every transcript, summary, and follow-up captured by Spinach becomes queryable context for your AI tools the moment the meeting ends.
The server supports structured queries across meeting types, attendees, dates, and topics, so an agent can pull Microsoft Teams meeting transcripts or the exact sprint planning discussion or architectural decision it needs. SOC2, GDPR, and HIPAA compliance means your conversations stay protected while your agents stay informed.
Final Thoughts on MCP Servers for Meeting Data
Setting up a meeting transcript MCP server means your agents stop working blind. They can query past decisions, pull action items, and reference architectural discussions without you manually feeding them context every time. Try Spinach’s MCP server to connect your meeting history to every coding agent you use. Your workflow gets faster when your AI knows what happened before you opened the editor.
Yes. Spinach provides a ready-to-use MCP server that connects to Claude, Cursor, Windsurf, and ChatGPT through a single OAuth setup. Authentication takes minutes, and your last 100 meetings become queryable context immediately—no custom infrastructure or maintenance required.
Meeting Baas requires self-hosting, manual API configuration, and ongoing maintenance as endpoints change. Spinach connects through OAuth with zero infrastructure setup and works across Claude, Cursor, Windsurf, and ChatGPT from one connection. If you need production-ready context without DevOps overhead, Spinach is the better fit.
An MCP (Model Context Protocol) server acts as a structured bridge that lets AI assistants and coding agents query your meeting data directly. Instead of manually copying transcript snippets into Claude or Cursor, the server feeds decisions, action items, and technical discussions as live context your agent can reason with.
Ask your AI agent a question that requires meeting data: “What were the key decisions from my last sprint review?” or “Find mentions of [feature name] across recent transcripts.” If the agent returns accurate information, your connection works. If it gives a generic “I don’t have access” response, check your config file for JSON syntax errors and verify your API credentials are valid.
Yes. Spinach’s MCP server lets agents query across your full meeting history, not just the most recent call. Your coding agent can track how a requirement evolved through multiple sprint reviews, surface past architectural decisions, or identify conflicting context across product planning calls—all without you reconstructing conversations from memory.
Yes. Spinach’s MCP server connects to Claude, Cursor, Windsurf, ChatGPT, and VS Code through a single OAuth connection. Configure your credentials once and the same meeting context flows across every agent you work with, so coding agents in Cursor can query decisions and action items just as easily as Claude Desktop.
Connect an MCP server that exposes your meeting transcripts directly to your AI assistant. Spinach’s MCP server gives Claude, Cursor, and other agents live access to your last 100 meetings through OAuth, so they can query decisions and technical discussions without you manually feeding them context every time.
No. While self-hosted options like Meeting Baas exist, Spinach provides a managed MCP server that connects through OAuth with no infrastructure setup or ongoing maintenance. Your meeting history stays accessible to Claude, Cursor, and other agents without you managing uptime or API changes.
Validate your config file with a JSON linter to catch syntax errors, then restart your MCP client completely. If tools still don’t appear, confirm your server is actually running by testing it locally first, and verify the command path in your config points to the correct entry file.
Building a custom MCP integration requires maintaining API auth, handling endpoint changes, and managing uptime as your meeting tool evolves. Spinach’s MCP server connects via OAuth in minutes and works across Claude, Cursor, Windsurf, and ChatGPT without custom code or infrastructure overhead.
Yes, if your MCP server supports multi-meeting queries. Spinach’s MCP server lets agents reason across your full meeting history, so they can track how a feature requirement changed through multiple sprint reviews or surface conflicting decisions from past planning calls without you reconstructing conversations.
You need Node.js 18 or higher. Run `node –version` in your terminal to check your current version, and grab the latest from nodejs.org if you’re running an older release.
Spinach’s MCP server connects to ChatGPT through the same OAuth flow you use for Claude and Cursor. Authenticate once in your Spinach settings and your last 100 meetings become queryable context in ChatGPT conversations without manual configuration or API key management.
MCP servers provide live, queryable access to your meeting history so agents can search by topic, attendee, or date without you manually selecting and uploading the right files. Spinach’s MCP server updates automatically after every meeting, so your agent always works with current context instead of stale uploads.
Yes. Spinach’s MCP server respects your account permissions and privacy settings, so agents only access meetings you’ve captured and have rights to view. You control what gets recorded and shared, and coding agents query only the transcript data you’ve already made available in your Spinach account.
What to do now
Next, here are some things you can do now that you've read this article:
- If communication is a challenge for your team, you should check out our library of meeting agenda templates.
- Check out Spinach to see how it can help you run a high performing org.
- If you found this article helpful, please share it with others on Linkedin or X (Twitter)
