> ## Documentation Index
> Fetch the complete documentation index at: https://docs.callecho.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Connect your AI app to Echo with MCP

> Connect an AI app to your Echo meetings, paste a starter prompt, and reuse requests for client calls, recruiting and travel consultations.

MCP lets your AI app read meeting context from Echo. Connect an app that supports remote MCP servers with OAuth sign-in, then ask it to find a meeting, check what someone said, or draft a follow-up from the evidence.

You approve each connection with your Echo account. Start with read-only access, or explicitly grant optional edits to your own Meeting notes, Contacts, and My Tasks. You can disconnect it at any time. For one URL to give an agent, use the [agent setup guide](https://docs.callecho.ai/anywhere/agent-setup.md).

## Let your AI agent help you get started

Paste this into the AI app where you want to use Echo. The agent can discover Echo's tools once the app is connected. Your app may need you to add the server in its settings first.

```text theme={null}
Set up Echo using this guide: https://mcp.callecho.ai/setup.md
```

If the agent says it cannot configure MCP, use the steps below. A prompt alone does not connect an app that lacks remote MCP support.

## Connect an app

1. In the AI app, add a remote MCP server or custom connector with this address. Choose Streamable HTTP if the app asks for a transport:

   ```text theme={null}
   https://mcp.callecho.ai/mcp
   ```

2. The app opens Echo's sign-in page. Sign in the way you sign in to Echo: Google, Microsoft or an email code. If your account uses an authenticator app, Echo asks for its code too.

3. Check the app's name and where it sends you back to, then select **Approve**.

Echo shows whether it could verify the app's name. An app that runs on your computer gets a warning, because any program on that computer could receive the access.

After approval, ask the agent to discover Echo's tools. Seeing an Echo entry in settings does not by itself confirm the connection works. If your app does not support remote MCP with OAuth, follow its official instructions or use [Echo Agent](https://docs.callecho.ai/agent/overview.md).

## Copy a request

Paste these requests into your connected AI app. Replace the names and details with your own. Northstar, Maya, Priya and the Novaks are fictional examples.

### General: catch up on a meeting

```text theme={null}
Use Echo MCP to find my most recent meeting with Northstar. Read its
notes and the relevant transcript passages. Summarize the decisions,
commitments and unresolved questions. Cite the meeting and any available
timestamps. Say which details you could not verify.
```

### General: draft a follow-up

```text theme={null}
Use Echo MCP to find my latest meeting with Maya at Northstar. Draft a
follow-up email using only the decisions and commitments in that meeting.
Keep my actions separate from Maya's. Flag missing owners or dates instead
of filling them in. Include the source meeting below the draft for review.
```

### Recruiting: prepare for an interview

```text theme={null}
Use Echo MCP to find my previous interviews with Priya Raman. Summarize
the evidence about her experience, motivation, compensation and availability.
Suggest follow-up questions for gaps or conflicting statements. Cite the
meetings and transcript passages. Ask me to choose if the name matches
more than one person. Do not infer a hiring decision.
```

### Recruiting: check a candidate against a role

```text theme={null}
Use Echo MCP to find my interviews with Priya Raman. Compare what she
actually said with the job description I provide in this conversation.
For each requirement, show supporting evidence or mark it as not established.
Keep claims separate from verified facts, cite the source meetings, and
list questions I still need to ask. Ask for the job description if it is missing.
```

### Travel advisor: recover a client's trip brief

```text theme={null}
Use Echo MCP to find my recent consultation with the Novaks about Japan.
Extract travelers, dates, destinations, budget, pace, accommodation preferences
and open questions. Cite the meeting and relevant passages. Separate explicit
preferences from your interpretation, and flag contradictions or missing details.
```

### Travel advisor: prepare a proposal draft

```text theme={null}
Use Echo MCP to read the Novaks' consultation about Japan. Draft a trip
proposal outline using the preferences and constraints they actually gave.
Cite the meeting context and mark missing information. Treat prices,
availability and entry requirements as details for me to confirm separately.
Keep the proposal as a draft for my review.
```

Your AI app writes these summaries and drafts from the meeting content Echo returns. For [candidate search](https://docs.callecho.ai/tracks/recruiting/find-candidates.md), [CV Out](https://docs.callecho.ai/tracks/recruiting/cv-out.md), connected-app actions or [scheduled tasks](https://docs.callecho.ai/agent/overview.md#schedule-something-for-later), use Echo Agent.

## Tools your agent can discover

A read-only general-purpose connection exposes four Meeting tools. The agent should use the names, descriptions and input schemas returned by tool discovery.

| Tool | Use it to |
| - | - |
| `recent_meetings` | Find recent meetings or meetings in a date range |
| `search_meetings` | Search meeting titles, notes and transcripts by text |
| `get_meeting` | Read a meeting using its returned ID; follow pagination for a long transcript |
| `search_transcripts` | Find transcript passages by meaning, with speakers and timestamps |

The optional `records:write` permission also exposes `read_echo_record`, `preview_echo_record_change`, and `apply_echo_record_change` for your own Meeting notes, Contacts, and My Tasks. It permits edits and completing or reopening existing Tasks, with revision checks and retry identity. It does not permit creation, deletion, sharing, Contact merges, privacy changes, or account settings. See the [edit workflow](https://docs.callecho.ai/anywhere/agent-setup.md#3-edit-only-with-delegated-permission). Existing connections need fresh consent to gain this permission.

Candidate search, CV Out, calendar actions, document Sources, and general Memory retrieval remain outside the general MCP connection. Other capabilities your AI app offers have their own permissions.

## What a connected app can do

* Read and search your meetings: titles, AI Notes and transcripts.
* Read meetings teammates shared with you, at the access they gave you. A share without the transcript stays without it.
* It never reads meetings marked **Exclude from automatic sharing** (Private — never auto-share in earlier versions).
* Without `records:write`, it cannot change anything in Echo. With that optional permission, it can make the narrow edits above on your behalf without another Echo approval for each edit.
* Private Meetings and Tasks derived from them are excluded from record editing. Shared Meeting read access never grants editing.

What the app reads leaves Echo and goes to that app and the AI provider it uses, under their terms.

## If the connection or a request fails

| What you see | What to try |
| - | - |
| The app has no remote MCP option | Check its current MCP support. Use Echo Agent if it cannot connect remote servers with OAuth |
| A sign-in or authorization error | Reconnect through the app and complete Echo's sign-in and approval steps |
| No matching meeting | Try the person's full name, company or meeting topic. Check that the meeting is saved and accessible; an empty result is not proof the conversation never happened |
| Notes arrive but no transcript | The meeting may have been shared without transcript access. Ask its owner for access if you need it |
| Echo asks you to wait | Follow the retry delay reported by the app; avoid repeatedly sending the same request |

See [Troubleshooting](https://docs.callecho.ai/reference/troubleshooting.md) for problems with Echo itself.

## Disconnect an app

Open **Settings › Integrations** and find **AI agents**. Select **Disconnect** next to the app. It stops working with Echo at its next request.

A connection also ends on its own after 30 days. The app then asks you to approve it again.

> [!NOTE]
>
> The AI agents list needs the latest Echo desktop version.

## Turn AI apps off for an organization

Owners and admins can stop everyone in an Organization from connecting AI apps: open **Settings › Organization** and turn off **Members can connect AI agents** under **External AI agents**. Turning it off ends every member's connections, including members who also belong to other Organizations. Members connect again after it's turned back on.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.