MedHook Documentation

Get a local engine running in three steps. Total time: under five minutes once Docker is installed.

git clone https://github.com/medhook/engine.git
cd engine
cp .env.example .env

docker compose up -d

# verify
curl http://localhost:3000/api/health
# → { "status": "healthy" }

# open the dashboard
open http://localhost:3000

Every integration in MedHook is a directed acyclic graph (DAG): a trigger fires, nodes execute, and edges connect them. Branching, parallel execution, and loops are all expressed as nodes — there is no separate orchestration layer.

Adapters are typed connectors — config validated by Zod, credentials encrypted with AES-256-GCM, connection-tested before use. Twelve adapters ship in the box.

import { z } from 'zod';
import { BaseAdapter } from '../base/adapter';

export class MyAdapter extends BaseAdapter {
  readonly type = 'my-adapter';
  readonly name = 'My Adapter';
  readonly description = 'Connects to my system';
  readonly version = '1.0.0';

  readonly configSchema = z.object({
    baseUrl: z.string().url(),
    apiKey: z.string().min(1),
  });

  readonly actions = [
    { id: 'read', label: 'Read records' },
    { id: 'write', label: 'Write records' },
  ];

  async read(profile, params) { /* ... */ }
  async write(profile, data)  { /* ... */ }
}

Every action in the engine is reachable through a REST API. JWT bearer tokens are obtained from /api/auth/token and expire after 2 hours. The full reference lives at engine/docs/API.md.

# Exchange your engine auth secret for a JWT
TOKEN=$(curl -s -X POST http://localhost:3000/api/auth/token \
  -H "Content-Type: application/json" \
  -d '{"secret":"'"$ENGINE_AUTH_SECRET"'"}' | jq -r .token)

# Use it
curl http://localhost:3000/api/workflows \
  -H "Authorization: Bearer $TOKEN"

curl -X POST http://localhost:3000/api/workflows \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "EHR → Lab Sync",
    "trigger": { "type": "schedule", "cron": "0 */6 * * *" },
    "nodes": [ /* DAG nodes */ ],
    "edges": [ /* DAG edges */ ]
  }'

Looking for the complete catalog (mapping sets, code tables, triggers, GraphQL, metrics)? See engine/docs/API.md.

MedHook is fully self-hosted. The fastest path is Docker Compose for local and single-server deployments; Terraform templates exist for AWS, Azure, and GCP.

# from the engine repo root
docker compose up -d

# tail logs
docker compose logs -f engine

# stop
docker compose down

cd engine/terraform/aws
terraform init
terraform apply

The engine exposes a Model Context Protocol stdio server so Claude Desktop (and other MCP clients) can drive your workflows directly. Twelve tools cover the workflow lifecycle, adapter management, and clinical-format parsing.

{
  "mcpServers": {
    "medhook": {
      "command": "node",
      "args": ["/absolute/path/to/engine/dist/mcp/server.js"],
      "env": {
        "MEDHOOK_ENGINE_URL": "http://localhost:3000",
        "MEDHOOK_AUTH_SECRET": "your-engine-auth-secret"
      }
    }
  }
}

Healthcare-grade defaults. PHI never leaves your infrastructure: medhook.dev only receives HMAC-signed usage analytics — never workflow data.