The Read AI MCP server is a remote Model Context Protocol server that gives AI assistants standardized access to your Read AI meeting data and capabilities. This article walks through adding it to an agent in Microsoft Copilot Studio as a custom MCP connection.
Once connected, your Copilot Studio agent can browse recent meetings, access your transcripts and other meeting details, deploy a Read AI meeting agent to record a Teams, Zoom, or Meet video call, and share your meeting reports with others.
Before you begin
You'll need:
- A Read AI account. You'll sign in with it during setup, and the agent will only be able to access meetings and reports that your account already has permission to see.
- Access to Microsoft Copilot Studio, with an agent already created (or permission to create one).
- Permission to add tools to that agent. In many organizations, adding a new connector requires an environment maker or admin role. If the Add a tool option isn't available to you, check with your Power Platform administrator.
Read AI MCP server details
You'll enter these values in the Copilot Studio setup wizard:
| Field | Value |
|---|---|
| Server URL | https://api.read.ai/mcp |
| Transport | Streamable HTTP |
| Authentication | OAuth 2.0 |
Copilot Studio supports the Streamable transport type, which is what the Read AI MCP server uses.
The Read AI server supports OAuth Dynamic Client Registration (DCR) with discovery, so you won't need to create a client ID, client secret, or paste in any authorization or token URLs. Copilot Studio's OAuth 2.0 → Dynamic discovery option finds and registers everything for you. Copilot Studio labels this option "OAuth 2.0"; it works with the Read AI server's OAuth 2.1 implementation.
Add the Read AI MCP server to your agent
We recommend using Copilot Studio's MCP onboarding wizard. It's the fastest path and doesn't require an OpenAPI schema file.
Step 1: Open the MCP onboarding wizard
- In Microsoft Copilot Studio, open the agent you want to connect to Read AI.
- Go to the Tools page for that agent.
- Select Add a tool.
- Select New tool.
- Select Model Context Protocol, which opens the MCP onboarding wizard.
Step 2: Enter the server details
Fill in the three required fields:
- Server name: for example, Read AI.
- Server description: a brief, clear description of what the server does. Your agent's orchestrator uses this at runtime to decide when to call Read AI, so make it specific. For example: Provides access to Read AI meeting recordings, transcripts, summaries, action items, and reports. Can also deploy a Read AI meeting agent to a scheduled meeting and share meeting reports with others.
-
Server URL:
https://api.read.ai/mcp
Step 3: Choose the authentication type
- For Authentication, select OAuth 2.0.
- For the OAuth 2.0 type, select Dynamic discovery.
You don't need to enter a client ID, client secret, authorization URL, or token URL. With dynamic discovery, Copilot Studio automatically discovers the necessary endpoints and registers itself with Read AI's identity provider.
Tip: If you see fields asking for an authorization URL or client credentials, you've selected Dynamic or Manual rather than Dynamic discovery. Go back and choose Dynamic discovery.
Step 4: Create the server and connect
- Select Create.
- Select Next to continue to the Add tool dialog.
- Choose Create a new connection for the Read AI MCP server.
- When prompted, sign in with your Read AI account and review the access being requested. Approve it to complete the connection.
- Select Add to agent.
The Read AI MCP server now appears on your agent's Tools page.
What your agent can do
Once connected, the following Read AI tools are available to your agent:
| Tool |
What it does
|
|---|---|
| Get meeting by ID | Retrieves full meeting details — metadata, participants, summary, chapter breakdown, action items, key questions, topics, full transcript with speaker identification, metrics, and recording download URLs. |
| List meetings | Browses recent meetings (up to 10 per page), with date-range filtering and pagination. |
| Create meeting agent | Deploys a Read AI bot to a Zoom, Google Meet, or Microsoft Teams meeting. Accepts a meeting URL, or the platform, meeting ID, and optional password. Supports optional scheduling and a custom title. |
| Share meeting report | Grants someone access to a specific meeting report at a chosen permission level (Viewer recap-only, Viewer full, or Editor), with an optional email invitation and personal message. |
Verify the connection
Test the agent before rolling it out:
- Open your agent's test pane in Copilot Studio.
- Ask something that requires Read AI data, for example, "List my most recent Read AI meetings."
- Confirm the agent calls the Read AI tool and returns your meetings.
If the agent answers without calling Read AI, sharpen the server description you entered in Step 2 so the orchestrator can tell when Read AI is the right source.
Troubleshooting
"Add a tool" or "Model Context Protocol" isn't available
Your account may not have permission to add tools in this environment. Contact your Power Platform or Copilot Studio administrator.
Sign-in fails or the connection won't authorize
Confirm you're signing in with the Read AI account that has access to the meetings you expect, and that pop-ups aren't blocked in your browser. Then delete the connection and create a new one from the Add tool dialog.
The connection is created, but the agent returns no meetings
The agent can only see what your Read AI account can see. Check that the signed-in account is a participant on, or has been granted access to, the meetings in question.
The agent isn't calling Read AI when it should
Rewrite the Server description to be more specific about the kinds of questions Read AI can answer. The orchestrator uses this description to route requests.
Your organization blocks the connection
Copilot Studio accesses MCP servers through Power Platform connectors, so any Power Platform data policy also governs access to the Read AI MCP server and its tools. If the connector is blocked, your administrator will need to adjust the applicable data loss prevention policy.
Related articles
- MCP Server — server details, compatible clients, and the full tool reference
- Add an existing MCP server to an agent — Microsoft's documentation, including the custom connector alternative
Still stuck? Reach out to Read AI support and we'll help you get connected.