RESTHeart

The Agent-Ready Backend for MongoDB

RESTHeart exposes your data through a native MCP server and full REST and GraphQL APIs, so AI agents and developers can read and write without custom integration code.

RESTHeart gives AI agents and developers instant, secure access to MongoDB. It ships with a native MCP server that any MCP-compatible client (Claude, Claude Code, Cursor, VS Code) can connect to out of the box, plus full REST, GraphQL, and WebSocket APIs for traditional application development. Built-in authentication and authorization, zero backend boilerplate, and a plugin framework for custom business logic when you need it.

Available as a Docker image and GraalVM native binary. Built on Java 25, Undertow, and virtual threads.

Core capabilities:

• MCP Server: Native Model Context Protocol server. Connect any MCP-compatible AI client (Claude, Claude Code, Cursor, VS Code) directly to your MongoDB data with one line of config.
• REST API: Full CRUD, aggregations, filtering, sorting, pagination
• GraphQL: Schema-driven mapping to MongoDB queries
• WebSocket: Real-time change streams and data sync
• SSE: Server-Sent Events for live dashboards, IoT feeds, and event streams
• Metrics & Monitoring: Built-in Prometheus-compatible metrics endpoint with a real-time browser dashboard at /metrics-ui — track request rates, latency percentiles, and HTTP status distribution
• IoT / MQTT: Connect devices and ingest telemetry directly into MongoDB (coming soon)
• Authentication and Authorization: JWT, OAuth2, LDAP, MongoDB-based users, ACL rules
• Plugin system: Extend with Java, Kotlin, JavaScript, or TypeScript for custom business logic

AI Agents & MCP

RESTHeart ships with Sophia, a native Model Context Protocol (MCP) server. Any MCP-compatible AI client can connect directly.

Public demo endpoints (no authentication required):

https://sophia-api.restheart.com/mcp/restheart/    — RESTHeart docs knowledge base
https://sophia-api.restheart.com/mcp/cloud/        — RESTHeart Cloud docs knowledge base

Connect with Claude Code:

claude mcp add --transport http sophia-restheart https://sophia-api.restheart.com/mcp/restheart
claude mcp add --transport http sophia-cloud https://sophia-api.restheart.com/mcp/cloud

Connect with Claude Desktop: Open Settings → Connectors → Add custom connector and paste one of the context URLs above.

Connect with Streamable HTTP clients (add to MCP settings):

{
  "sophia": {
    "type": "http",
    "url": "https://sophia-api.restheart.com/mcp/restheart/"
  }
}

Connect with stdio clients (VS Code, Zed — requires Node.js 18+):

{
  "sophia": {
    "command": "npx",
    "args": ["mcp-remote", "https://sophia-api.restheart.com/mcp/restheart/"]
  }
}

For private Sophia instances, replace the URL with your own base URL and context ID, and add authentication. See the MCP documentation for details.

Quick Start

To get RESTHeart running in few minutes and make your first API call, there are several options. Choose the one that best fits your needs:

Option 1 — Deploy on RESTHeart Cloud (no install required)

The fully managed service by SoftInstigate. No infrastructure to configure or maintain — just sign up and start building. Includes a dashboard, SLA, and support.

Option 2 — Deploy on Railway (no install required)

Deploy RESTHeart and MongoDB on your own Railway account in one click. You keep full control of the infrastructure. Free to start, pay only for what you use on Railway.

Option 3 — Docker Compose (local):

# Start MongoDB + RESTHeart with Docker Compose
curl https://raw.githubusercontent.com/SoftInstigate/restheart/master/docker-compose.yml \
  --output docker-compose.yml && docker compose up --attach restheart

# Test it
curl http://localhost:8080/ping

Default credentials: admin / secret (change in production)

More options: https://restheart.org/docs/foundations/quick-start

Example: Query MongoDB via HTTP

const url = encodeURI('https://demo.restheart.org/messages?filter={"from":"Bob"}&pagesize=1');

fetch(url)
  .then(response => response.json())
  .then(json => console.log(JSON.stringify(json, null, 2)));

That's it. No Express routes, no Mongoose schemas, no middleware setup.

📄 Full documentation: https://restheart.org/docs/

Use Cases

• AI agent backends: Give your agents a production-grade MongoDB backend via MCP. No custom integration code.
• API development without boilerplate: Skip CRUD code, focus on business logic
• Mobile and web backends: Get REST/GraphQL APIs immediately
• Real-time applications: WebSocket and SSE for chat, notifications, and live dashboards
• IoT backends: Collect and store sensor data via MQTT, query it via REST (coming soon)
• MongoDB Data API replacement: Self-hosted alternative to the deprecated Atlas Data API (migration guide)
• Legacy modernization: Add modern APIs to existing MongoDB databases
• PostgreSQL with MongoDB API: Use via FerretDB for PostgreSQL storage (tutorial)

Metrics & Monitoring

RESTHeart ships with a built-in real-time metrics dashboard and a Prometheus-compatible endpoint. No external tools required.

Metrics UI — Open /metrics-ui in your browser after starting RESTHeart to access the real-time dashboard.

The dashboard displays:

• KPIs: total requests, request rate (1-min / 5-min), latency percentiles (p50, p95, p99)
• Time-series chart: request rate over time per method+status
• Bar chart: cumulative request counts per method+status
• Latency distribution: p50, p75, p95, p98, p99, p999 breakdowns

Prometheus endpoint — Scrape /metrics (or /metrics/requests, /metrics/jvm) in Prometheus format for integration with external monitoring stacks.

Configuration — Enabled by default. Key settings in restheart.yml:

metrics:
  enabled: true
  uri: /metrics

requestsMetricsCollector:
  enabled: true
  include: ["/*"]
  exclude: ["/metrics", "/metrics/*"]

jvmMetricsCollector:
  enabled: false   # set to true for JVM memory and GC metrics

📖 Full documentation: docs/metrics.md

Extend with Plugins

Write custom logic only when you need it. RESTHeart handles the rest.

Java Plugin

@RegisterPlugin(name = "greetings")
public class GreeterService implements JsonService {
    @Override
    public void handle(JsonRequest req, JsonResponse res) {
        res.setContent(object()
            .put("message", "Hello World!")
            .put("timestamp", Instant.now()));
    }
}

JavaScript Plugin

export const options = {
    name: "greetings",
    uri: "/greetings"
}

export function handle(request, response) {
    response.setContent(JSON.stringify({
        message: 'Hello World!',
        timestamp: new Date().toISOString()
    }));
    response.setContentTypeAsJson();
}

Plugin types:

• Services: custom REST endpoints
• Interceptors: modify requests and responses, add validation
• Initializers: run code at startup
• Providers: dependency injection

📖 Plugin development: https://restheart.org/docs/plugins/overview/

🔧 Use restheart-cli for scaffolding, testing, and hot-reload.

Deployment

Docker

docker pull softinstigate/restheart:latest

docker run -p 8080:8080 \
  -v ./restheart.yml:/opt/restheart/etc/restheart.yml \
  softinstigate/restheart

Kubernetes

Stateless architecture supports horizontal scaling. Configure with ConfigMaps and Secrets.

Native Executables

Prebuilt binaries for macOS, Linux, and Windows with faster startup and lower memory footprint.

See docs/native-executables.md for download links.

RESTHeart Cloud

Fully managed service: cloud.restheart.com

• Instant provisioning
• Automatic scaling
• Free tier available
• Premium plugins (Webhooks, Sophia AI, Facet)

Database Compatibility