Read AI now offers both a REST API and an MCP (Model Context Protocol) server, enabling programmatic access to meeting data such as transcripts, summaries, and action items. In addition to direct API integration, users can connect third-party AI applications and platforms (such as ChatGPT, Claude, and other LLM-powered tools) as MCP clients, allowing these systems to query and interact with their Read AI meeting data seamlessly.
These capabilities are currently available as part of an open beta. We encourage you to explore both the API and MCP server to see whether they will fit your integration needs.
Current State
- Open Beta: Both the API and MCP server are in open beta, accessible to all users who meet the prerequisites described below. There are also some known limitations outlined further below, with improvements coming soon.
- Core Capabilities: The API and MCP server allow you to retrieve meeting reports, live transcripts, summaries, and more. Both provide access to the same data but through different means.
- Documentation:
Prerequisites
-
If you belong to a workspace, it must have the Downloads option enabled in Workspace Settings > Reports & Sharing:
- It is otherwise available to all users regardless of plan or workspace.
Known Issues and Limitations
- Authentication Complexity: The current OAuth 2.1 flow requires a browser-based login and does not yet support static API keys or client credentials. Access tokens expire after 10 minutes, and refresh tokens are single-use and rotate on each refresh (with a short grace period). This can introduce friction for persistent or machine-to-machine integrations, and may require manual intervention if token chains break.
- SSO Redirect: Users can authenticate using SSO, however there have been some reports of users not getting properly redirected back into the OAuth flow after signing in with your organization's SSO (SAML). Users can restart the OAuth process after signing in to work around this.
- Rate Limits: The API enforces a rate limit of 100 requests per minute per user. Exceeding this limit will result in HTTP 429 errors.
- Feature Gaps: Some features available in the Read AI web app (such as Ask Read, Coaching, and other integration features) are not yet exposed via API/MCP. Uploading audio or video files for transcription via API is also not currently supported.
- Live Transcript: While live transcription is available via the API, by default it is not enabled for a meeting unless someone accesses the live dashboard. A setting is planned to enable this for all meetings in your account.
- Cross-Platform Interoperability: There have been reports of issues with some external MCP clients (e.g., VS Code, Notion) not working with our authentication, and broader compatibility testing is ongoing.
- Permissions: Users can only retrieve data for meeting reports that they have access to in the Read AI web app. Admins must enable global report access in order to access all reports within a workspace.
Feedback and Improvements
We are actively collecting feedback to improve the developer experience and expand capabilities. Planned enhancements for GA include:
- Support for static API keys/personal access tokens
- More options for token lifetimes and a refresh token grace period
- A setting to enable live transcription for all meetings
- Additional endpoints, tools, and webhook/event support
- Expanded documentation and self-serve UI for easier onboarding
If you encounter any issues, have suggestions, or require assistance, please contact our support team.
Continue Reading
For detailed setup instructions, API reference, and MCP usage guides, please visit: