Sayli as an MCP server

Read your meetings from your own AI tools. Claude, Cursor, ChatGPT, all read-only.

Sayli can be a source your own AI tools read from. Connect it as an MCP server and your meetings become something Claude, Cursor, or ChatGPT can ask about, without copying anything out of Sayli. It speaks JSON-RPC, authenticates with a personal access token, and is strictly read-only.

The six tools

ask_meetings

Ask a question across your meetings and get a grounded answer.

search_meetings

Search for moments across every call you can see.

list_meetings

List your recent meetings.

get_brief

Pull the brief for one meeting.

list_my_actions

List the action items that are on you.

list_folders

List your folders.

Connect it

Mint a connection

In Settings, connect the MCP server. Sayli mints an org-bound token and gives you a paste-ready client config.

Paste it into your tool

Add the config to your AI client. Sayli returns it in the mcpServers shape Claude Desktop and Cursor read:

json

{
  "mcpServers": {
    "sayli": {
      "url": "https://api.sayli.ai/api/v1/mcp",
      "headers": { "Authorization": "Bearer sayli_pat_your_token_here" }
    }
  }
}

Clients configured through a UI, like ChatGPT connectors, do not take this JSON. Paste the URL and the token into the connector form instead.

Read-only and audited

The server can read your meetings. It cannot change anything. Every call is written to your audit log and fires an mcp.tool_called webhook, so you always know what was asked and when.

Personal access tokens

Create the token your client uses to connect.

The same memory, searched from inside the app.