Agent access

Let your AI read the meetings.

Ledgeur speaks the Model Context Protocol, so an agent can list your meetings, search them, read a full transcript with speakers, and pull the open action items — without you pasting anything.

The tools

Four tools, generated from the code that implements them.

This list is built from the real tool definitions at build time, so it cannot describe something that does not exist.

list_meetings

List the user's most recent meetings (title, date, status). Use to browse before drilling into one.

search_meetings

Search the user's meetings by keyword in the title. Returns matching meetings.

get_meeting

Get a single meeting with its notes, speakers and full transcript by meeting id.

list_tasks

List action items (tasks) extracted from meetings, optionally filtered by status.

Connecting

Generate a token on your account page, then give your MCP client one of these. The hosted endpoint is the one to prefer: there is nothing to install, and it works from a phone, a hosted agent or someone else’s machine.

Hosted — nothing to run

{
  "mcpServers": {
    "ledgeur": {
      "type": "http",
      "url": "https://ledgeur.com/api/mcp",
      "headers": {
        "Authorization": "Bearer ldg_your_token_here"
      }
    }
  }
}

Or run the server yourself

For clients that only speak stdio. This talks to the same database and exposes the same tools.

npx -y @ledgeur/mcp-server

What an agent can see

Exactly what you can see, and nothing else. A token is exchanged for a short-lived session belonging to you, and every query runs under the database’s row-level security. An agent cannot read a meeting you could not read, and it cannot read another workspace’s meetings even if it asks.

It also cannot see anything you have not synced. Meetings on the free plan live only in your browser, so there is nothing for an agent to connect to — which is the trade being made when you pay: the record becomes reachable, deliberately.

What it cannot do

Write. Every tool is read-only. An agent cannot edit a transcript, rename a speaker, or delete a meeting.
Hear anything. Audio is never synced, so there is no recording for an agent to reach.
Identify a voice. Voice prints stay on the device that made them and are excluded from sync.

Things worth asking it

“What did we decide about pricing, and when?”
“What am I supposed to be doing after this week’s meetings?”
“Has this customer complaint come up before?”
“Summarise everything said about the migration, in order.”

Part of the Team plan

$6 per person, per month.

Agent access comes with sync and the shared team library. Fourteen days free, and the local app stays free whatever you decide.

See pricing Generate a token