Product Updates
Fedor Zhilkin
Jul 9, 2026
·
Updated on
Jul 9, 2026
Mymeet.ai supports MCP — an open protocol for connecting AI agents to external data sources. Once connected, Claude, Cursor, or another AI tool gains direct access to your meetings: transcripts, summaries, tasks, and participants. Ask the agent about a past meeting — it finds the data on its own, no manual copying required.
In this article — what MCP is, two connection methods, step-by-step instructions for specific agents, and a full overview of what AI can do with Mymeet.ai data.
What MCP Is and How It Works in Mymeet.ai
MCP stands for Model Context Protocol — an open standard for connecting AI agents to external systems. Developed by Anthropic (the creators of Claude) and supported by most modern AI tools. Instead of manually copying data into a chat with the agent, you configure the connection once — and the agent retrieves the information it needs on demand.
In the context of Mymeet.ai, it works like this: you ask Claude "what did we discuss at the team meeting on Tuesday?" — the agent queries Mymeet.ai, finds the relevant meeting, and responds based on the actual transcript data.
Why Use MCP When There's Already a REST API?
The API and MCP provide access to the same data, but through different interfaces. The API requires writing code: you define the request logic, handle responses, and build the integration. It's flexible, but requires technical knowledge.
MCP works through dialogue: you write a request in plain language — the agent figures out what data to query and how to present it. No code needed. If your goal is to work with meeting data through an AI assistant in everyday work, MCP is simpler and faster.
What an AI Agent Can Do via Mymeet.ai MCP
After connecting, the agent gets access to 11 tools for working with meeting data. This isn't just read-only access — the agent can search across meetings, retrieve specific sections of reports, and even trigger recording of new meetings.
Full list of capabilities:
Get a list of recent meetings with filtering by date and participants
Check the processing status of a meeting
Retrieve an AI summary with key moments and tasks
Read the full transcript with timestamps
Find meetings by name, participant, or keyword
Download a report in Markdown, JSON, PDF, or DOCX format
Launch a meeting recording via bot
Rename a meeting or apply a different analysis template
Delete a meeting
Through MCP, the agent works with data from the workspace the API key is linked to. Meetings from other workspaces are inaccessible until the workspace is changed in settings.
Two Ways to Connect Mymeet.ai MCP
Mymeet.ai supports two MCP connection methods — local and remote. They differ in where the server runs that handles communication between the agent and Mymeet.ai.
Local Connection via npx
The server runs on your computer. This is the primary method for most desktop AI tools: Claude Desktop, Claude Code, Cursor. Node.js version 18 or higher must be installed (Node.js is a runtime for executing JavaScript code outside a browser — free to download from nodejs.org).
You don't need to start the server manually — the AI tool does this automatically on each launch using npx. npx is a tool that downloads and runs a package (in this case, the Mymeet.ai MCP server) without permanently installing it on your computer.
The configuration looks the same for all tools using local connection:
{
"mcpServers": {
"mymeet": {
"command": "npx",
"args": ["-y", "@mymeet/mcp-server"],
"env": {
"MYMEET_API_KEY": "your_key_from_settings"
}
}
}
}
What each part means:
"command": "npx"— the command to launch the server"args": ["-y", "@mymeet/mcp-server"]— the name of the Mymeet.ai MCP server package;-ymeans auto-confirm installation"MYMEET_API_KEY"— your API key from Mymeet.ai settings
This config is pasted into the settings file of your chosen AI tool — the exact location for each is shown below.
Remote Connection via HTTP
The server is already running on Mymeet.ai's side — nothing needs to be installed on your computer. The connection goes directly over the internet at https://mcp.mymeet.ai/mcp. This option is suited for browser-based tools and team deployments where one config is needed for multiple users.
Config for remote connection:
{
"mcpServers": {
"mymeet": {
"url": "https://mcp.mymeet.ai/mcp",
"headers": {
"Authorization": "Bearer your_key_from_settings"
}
}
}
}
What each part means:
"url"— the address of the Mymeet.ai MCP server"Authorization": "Bearer ..."— the authorization header; Bearer is the standard token type in HTTP authorization, followed by your API key
How to Connect Specific AI Agents to Mymeet.ai via MCP
The config is the same for all agents — the only difference is where to paste it. Instructions for each tool are below.
Claude Code — Connecting Mymeet.ai MCP
Claude Code is an AI assistant for development that runs in the terminal. Connecting via MCP here is the most straightforward: one config file, and the agent immediately gains access to meeting data within the working session.
Open or create the file .claude/mcp.json in your home directory. Paste the local connection config with your API key:
{
"mcpServers": {
"mymeet": {
"command": "npx",
"args": ["-y", "@mymeet/mcp-server"],
"env": {
"MYMEET_API_KEY": "your_key"
}
}
}
}
After saving the file, restart Claude Code — the Mymeet.ai server will appear in the list of connected tools. You can now ask the agent about meetings directly in the terminal: "show the summary of the last meeting," "what did the team discuss last week."
Claude Desktop — Connecting Mymeet.ai MCP
Claude Desktop is Anthropic's desktop application for macOS and Windows. Open Claude Desktop settings and go to Developer → Edit Config. The claude_desktop_config.json file will open — add the mcpServers block with the Mymeet.ai config:
{
"mcpServers": {
"mymeet": {
"command": "npx",
"args": ["-y", "@mymeet/mcp-server"],
"env": {
"MYMEET_API_KEY": "your_key"
}
}
}
}
Save the file and restart Claude Desktop. A tools icon will appear in the interface — this means MCP is connected and Claude can now access Mymeet.ai data.
Cursor — Connecting Mymeet.ai MCP
Cursor is an AI code editor with MCP support. Open Cursor settings: Cursor Settings → MCP → Add new MCP server. Paste the local connection config with your key and save. Cursor restarts the server automatically — after that, the AI assistant in the editor gains access to Mymeet.ai meeting data.
Other MCP-Compatible Tools
Any AI tool with MCP support connects using one of the two configs above — local or remote. For browser-based tools and tools without a local runtime, use the remote connection via https://mcp.mymeet.ai/mcp.
Example Requests to an AI Agent via Mymeet.ai MCP
After connecting, the agent understands requests in plain language — no commands or formats to learn. Just describe what you want the way you'd ask a colleague.
A few examples of how you can work with meeting data:
"Show my recent meetings from this week"
"What did we discuss on the client call on Tuesday?"
"Find meetings that included Ivan Petrov"
"What tasks were assigned at yesterday's sync?"
"Download the report for the 'Quarterly Planning' meeting in Markdown"
"Record the Google Meet meeting at meet.google.com/xxx tomorrow at 3 PM"
The agent determines on its own what data to request and responds with real data from Mymeet.ai — without switching between applications.
Mymeet.ai MCP Plans and Limits
The MCP server uses the same API key and inherits the limits of the workspace plan the key is linked to. On the Free plan, AI agent cards are visible in the Integrations section, but MCP connections do not work.
Plan | MCP Connection | Available Data |
Free | Unavailable | — |
Lite | Full access | Meetings in the selected workspace |
Pro | Full access | Meetings in the selected workspace |
Business | Full access | Meetings in the selected workspace |
Switch the workspace using the selector on the API key page — the agent will immediately start seeing meetings from the new workspace without reconfiguring the config.
Summary
The Mymeet.ai MCP integration lets you work with meeting data through an AI agent without writing code or manually copying transcripts. Add the config to Claude, Cursor, or another tool's settings once — and the agent gains access to all meetings in the workspace.
If you need programmatic integration and automation through code — read the Mymeet.ai REST API guide.
Frequently Asked Questions About Mymeet.ai MCP
What is MCP and why does Mymeet.ai need it?
MCP (Model Context Protocol) is an open standard for connecting AI agents to external systems. In Mymeet.ai, it allows Claude, Cursor, and other AI tools to access meeting data directly — transcripts, summaries, tasks — without manual copying.
Do I need to install anything extra for Mymeet.ai MCP?
For local connection, you need Node.js 18+ — a code execution runtime. The Mymeet.ai MCP server itself is downloaded automatically on first launch via npx. For remote connection via https://mcp.mymeet.ai/mcp, nothing needs to be installed.
Does Mymeet.ai MCP work in the browser versions of Claude or ChatGPT?
Local mode (via npx) only works in desktop applications: Claude Desktop, Claude Code, Cursor. For browser-based tools, use the remote connection via URL https://mcp.mymeet.ai/mcp, if the tool supports it.
Which AI agent is easiest to connect to Mymeet.ai via MCP?
Claude Code — there's no graphical settings UI; the config is added to a single file and works immediately. Claude Desktop is also straightforward — through the built-in config editor in the application settings.
What data does the AI agent see after connecting via MCP?
The agent sees meetings from the workspace selected via the selector on the API key page: transcripts, summaries, participants, tasks, and decisions. Meetings from other workspaces are inaccessible until the workspace is changed in settings.
Do I need to reconfigure MCP when switching workspaces?
No. The agent config and key stay the same. Simply change the workspace using the selector on the API key page — the agent will start seeing meetings from the new workspace.
Which plans support Mymeet.ai MCP?
Paid plans: Lite, Pro, and Business. On the Free plan, the AI Agents section in Integrations is available for exploration, but connections don't work.
Can an AI agent via MCP launch a meeting recording, not just read data?
Yes. The Mymeet.ai MCP server includes a recording tool: the agent can send a bot to a meeting by providing the link, platform, and time. This is one of the 11 operations available through MCP.
What should I do if the AI agent doesn't see meetings after connecting MCP?
Check three things: whether the API key in the config is correct, which workspace the key is linked to on the settings page, and whether a paid plan is active on that workspace. On the Free plan, MCP requests do not execute.
Where can I find the full Mymeet.ai MCP server documentation?
Full documentation with descriptions of all tools, parameters, and additional config examples: https://github.com/MyMeetAI/mymeet-mcp-server/blob/main/README.ru.md
Fedor Zhilkin
Jul 9, 2026
