# Health Records MCP Server for AI Agents

Flexpa's patient records MCP server at https://api.flexpa.com/mcp/consent lets AI agents query health records authorized through Flexpa Consent. It uses patient access tokens and provides tools such as get_health_summary, get_patient, get_coverage, and get_claims, plus read-only resources such as flexpa://medications and flexpa://labs.

How it works: Flexpa Consent -> Patient access token -> MCP tools and resources. Application access tokens do not grant access to patient records tools.

## Connect

### Claude Code

```bash
claude mcp add flexpa \
  --transport http \
  https://api.flexpa.com/mcp/consent
```

### Tool call

```json
{
  "name": "get_health_summary",
  "arguments": { "sections": ["problems", "medications"] }
}
```

## Good to know

- Authentication uses OAuth 2.1 with PKCE through Flexpa Consent.
- Application access tokens don't grant access to patient records tools.
- The public registry of health tools is at /mcp, with no authentication.

## Implementation status

Flexpa documents the patient records MCP server as beta. Check the agents guide for current tools and configuration.

## Questions

**Is there an MCP server for health records?** Yes. Flexpa's MCP server at https://api.flexpa.com/mcp/consent gives agents patient-consented records through tools and resources.

**How does authentication work?** Flexpa's patient records MCP server uses OAuth 2.1 with PKCE through Flexpa Consent and accepts the same patient access tokens as the FHIR API. Application access tokens do not grant access to patient records tools.

## More

- [Health records for agents (docs)](https://www.flexpa.com/docs/guides/agents)
- [Health tools for agents](https://www.flexpa.com/agents)

Last updated 2026-11-11.