Skip to content
MCP server

Last EHR — Medplum

Read-only FHIR chart tools for MCP, with human-approved writes

Runs locallyStatus and history
Install
Pricing
Open source
Platforms
CLI, self-hosted
Runs
Locally
Version
0.5.4
Listing updated
3 days ago

About Last EHR — Medplum

Last EHR — Medplum is an open-source MCP server for health and science. It runs locally on your machine and is started by your AI client.

The notable choice here is that writes are opt-in and gated by a person. By default the package offers four read tools: patient search, chart overview, a filtered chart section covering 23 patient-scoped sections that report incomplete or unsupported reads, and inline document text. Setting LASTEHR_MCP_WRITES=proposal adds four write tools for notes (Communication), measurements and linked corrections (Observation) and follow-up tasks (Task). Each needs a human approval prompt, so a client without MCP elicitation support gets no write tools. Existing records are never updated or deleted.

It works with hosted or self-hosted Medplum, a no-auth HAPI stack for local synthetic testing, and an in-memory synthetic store started with init --demo that needs no account or Docker. Experimental adapters reach the Epic and Oracle Health sandboxes and generic SMART Backend Services, which the project says are not a supported production setup.

The server returns chart data to your MCP client, which may pass it to its model provider; the package itself calls no model. The project states that its notes are not a clinical or compliance claim and suggests starting with synthetic records.

Install

Run in a terminal

claude mcp add last-ehr --env MEDPLUM_ACCESS_TOKEN='<MEDPLUM_ACCESS_TOKEN>' --env MEDPLUM_CLIENT_SECRET='<MEDPLUM_CLIENT_SECRET>' --env LASTEHR_PRIVATE_KEY='<LASTEHR_PRIVATE_KEY>' -- npx -y @lastehr/mcp

Add to claude_desktop_config.json

{
  "mcpServers": {
    "last-ehr": {
      "command": "npx",
      "args": [
        "-y",
        "@lastehr/mcp"
      ],
      "env": {
        "MEDPLUM_ACCESS_TOKEN": "<MEDPLUM_ACCESS_TOKEN>",
        "MEDPLUM_CLIENT_SECRET": "<MEDPLUM_CLIENT_SECRET>",
        "LASTEHR_PRIVATE_KEY": "<LASTEHR_PRIVATE_KEY>"
      }
    }
  }
}

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "last-ehr": {
      "command": "npx",
      "args": [
        "-y",
        "@lastehr/mcp"
      ],
      "env": {
        "MEDPLUM_ACCESS_TOKEN": "<MEDPLUM_ACCESS_TOKEN>",
        "MEDPLUM_CLIENT_SECRET": "<MEDPLUM_CLIENT_SECRET>",
        "LASTEHR_PRIVATE_KEY": "<LASTEHR_PRIVATE_KEY>"
      }
    }
  }
}

Add to .vscode/mcp.json

{
  "servers": {
    "last-ehr": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@lastehr/mcp"
      ],
      "env": {
        "MEDPLUM_ACCESS_TOKEN": "<MEDPLUM_ACCESS_TOKEN>",
        "MEDPLUM_CLIENT_SECRET": "<MEDPLUM_CLIENT_SECRET>",
        "LASTEHR_PRIVATE_KEY": "<LASTEHR_PRIVATE_KEY>"
      }
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "last-ehr": {
      "command": "npx",
      "args": [
        "-y",
        "@lastehr/mcp"
      ],
      "env": {
        "MEDPLUM_ACCESS_TOKEN": "<MEDPLUM_ACCESS_TOKEN>",
        "MEDPLUM_CLIENT_SECRET": "<MEDPLUM_CLIENT_SECRET>",
        "LASTEHR_PRIVATE_KEY": "<LASTEHR_PRIVATE_KEY>"
      }
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "last-ehr": {
      "command": "npx",
      "args": [
        "-y",
        "@lastehr/mcp"
      ],
      "env": {
        "MEDPLUM_ACCESS_TOKEN": "<MEDPLUM_ACCESS_TOKEN>",
        "MEDPLUM_CLIENT_SECRET": "<MEDPLUM_CLIENT_SECRET>",
        "LASTEHR_PRIVATE_KEY": "<LASTEHR_PRIVATE_KEY>"
      }
    }
  }
}

Configuration

  • MEDPLUM_ACCESS_TOKENFHIR chart tools over Medplum or a local HAPI stack: read-only by default, with an opt-in elicitation-gated write profile (human-approved per action).
  • MEDPLUM_CLIENT_IDFHIR chart tools over Medplum or a local HAPI stack: read-only by default, with an opt-in elicitation-gated write profile (human-approved per action).
  • MEDPLUM_CLIENT_SECRETFHIR chart tools over Medplum or a local HAPI stack: read-only by default, with an opt-in elicitation-gated write profile (human-approved per action).
  • MEDPLUM_BASE_URLFHIR chart tools over Medplum or a local HAPI stack: read-only by default, with an opt-in elicitation-gated write profile (human-approved per action).
  • LASTEHR_BACKENDepic, oracle, or smart. Requires LASTEHR_EXPERIMENTAL_FHIR_ADAPTERS=1. Sandbox writes are approval-gated and persist because those sandboxes do not support delete.
  • LASTEHR_EXPERIMENTAL_FHIR_ADAPTERSSet to 1 to enable LASTEHR_BACKEND epic, oracle, or smart.
  • LASTEHR_CLIENT_IDSMART Backend Services client id. Not a secret.
  • LASTEHR_PRIVATE_KEYPKCS#8 PEM for SMART private_key_jwt. Never log it.
  • LASTEHR_PRIVATE_KEY_PATHPath to the PKCS#8 PEM. Alternative to LASTEHR_PRIVATE_KEY.
  • LASTEHR_KEY_IDJWT kid. Must match the public JWKS.
  • LASTEHR_JWKS_URLhttps URL of the public JWKS. doctor checks that LASTEHR_KEY_ID is present. Optional.
  • LASTEHR_SCOPESSpace-delimited SMART scopes. Required for Oracle. Omit for Epic unless overriding the app registration.
  • LASTEHR_FHIR_BASE_URLFHIR base URL. Epic and Oracle default to their public sandboxes when unset.
  • LASTEHR_TOKEN_URLSMART token URL. Epic and Oracle default to their public sandboxes when unset.
  • LASTEHR_ENCOUNTEREncounter/{id} added to Epic observations when the tool call omits encounter.
  • LASTEHR_PERFORMERPractitioner/{id} filed as legal authenticator on Oracle observations.

What it does

Four read tools
Patient search, chart overview, filtered section and inline document text.
Approval-gated writes
Notes, observations and tasks need a human approval prompt.
Demo mode
init --demo runs synthetic patients in-process with no account.
Medplum and HAPI
Works with Medplum projects and a local HAPI evaluation stack.
SMART sandboxes
Experimental Epic and Oracle Health sandbox adapters.
23 chart sections
The section reader covers 23 patient-scoped sections.

Good for

  • Prototype a chart-summary agent on synthetic patients
  • Let an agent draft a note that a clinician approves
  • Explore a FHIR record set through natural language
  • Test SMART Backend Services against vendor sandboxes

Pricing

Open source

Open-source package

Prices change. Check lastehr.com before you buy.

Last EHR — Medplum FAQ

Are writes enabled by default?

No. It is read-only unless LASTEHR_MCP_WRITES=proposal is set, and each write needs approval.

Where does patient data go?

It goes from the FHIR backend to your MCP client, which may send it to its model provider.

Can I try it without a backend?

Yes, init --demo uses a synthetic in-memory store.

Which backends work?

Medplum, a local HAPI stack, and experimental Epic, Oracle Health and SMART sandbox adapters.

Is Last EHR — Medplum free?

Yes. Last EHR — Medplum is open source, so you can use and self-host it for free.

How do I add Last EHR — Medplum to Claude Code?

Run this command in your terminal: claude mcp add last-ehr --env MEDPLUM_ACCESS_TOKEN='<MEDPLUM_ACCESS_TOKEN>' --env MEDPLUM_CLIENT_SECRET='<MEDPLUM_CLIENT_SECRET>' --env LASTEHR_PRIVATE_KEY='<LASTEHR_PRIVATE_KEY>' -- npx -y @lastehr/mcp

What are the best alternatives to Last EHR — Medplum?

The closest alternatives to Last EHR — Medplum are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.

Alternatives to Last EHR — Medplum

  1. REA

    MCP server

    CLI and MCP server that lets coding agents reverse engineer apps and binaries

    Runs locally
    • Open source
    • 464K GitHub stars
    • Developer tools
  2. Codebase Memory

    MCP server

    Local MCP server that builds a code knowledge graph for AI coding agents

    Runs locally
    • Open source
    • 48.7K GitHub stars
    • Knowledge and memory
  3. GitHub

    MCP server

    Official MCP server connecting AI assistants to GitHub repos, issues and workflows

    Not checked yet
    • Remote
    • Open source
    • 34.5K GitHub stars
    • Developer tools
  4. MCP toolkit giving coding agents symbol-level code retrieval and editing

    Runs locally
    • Open source
    • 31.5K GitHub stars
    • Developer tools
  5. XERJ

    MCP server

    Local search engine for AI agents with cited answers over your code and docs

    Runs locally
    • Open source
    • 19.2K GitHub stars
    • Knowledge and memory
  6. Ouroboros

    MCP server

    Spec-first layer that pins requirements and verifies results of AI coding agent work

    Runs locally
    • Open source
    • 6.4K GitHub stars
    • Developer tools