Give your coding agenta real backend.

Database, REST API, auth and row policies — built by your AI editor over MCP, reviewable by you in one dashboard.

Start building for freeConnect your agent

One project free · No credit card · Your data stays portable

New to this? Read what an MCP backend is, how an MCP database differs from a database MCP server, how to connect a remote MCP server, or how MCP OAuth sign-in works.

SQLiteREST APIAuthRow policiesWebhooksRemote MCPOpenAPIEject

Everything you needbehind your next app

Your agent writes the app. MCPBackend gives it structured tools for the data and the access rules behind it, plus signed webhooks when rows change.

Database tables

Tables, relationships and indexes on SQLite — defined by your agent, visible to you.

Row-level access

Public, signed-in or owner-only, set per table and enforced by the API.

Instant REST API

CRUD endpoints with filters, sorting and pagination, plus an OpenAPI contract.

End-user auth

Email and password accounts for your app's users, with tokens for the browser.

Portable by design

Export your SQLite data and a compatible runtime whenever you want to self-host.

See the backendin action

Click through the calls your agent makes — and the API your app gets back. Illustrative example project.

Tables your agent can reason about

Columns, types, constraints and indexes, created through MCP and visible in the dashboard.

ColumnTypeConstraint
idINTEGERPrimary key
titleTEXTRequired
statusTEXTDefault 'todo'
owner_idTEXTOwner access
RequestMCP
tool create_table(
  name: "tasks",
  columns: ["title TEXT NOT NULL",
            "status TEXT", "owner_id TEXT"]
)
Response{ "table": "tasks", "columns": 4, "indexes": 1 }

Built to pluginto anything

Let your agent configure it over MCP, call it from any app over REST, or get pushed the moment a row changes.

Remote MCP server
// .mcp.json
{
  "mcpServers": {
    "mcpbackend": {
      "type": "http",
      "url": "https://mcp.mcpbackend.com/mcp"
    }
  }
}

Ship productsnot infrastructure

Skip the backend boilerplate. Keep the parts that matter visible and in your control.

Agent-operable

Structured MCP tools instead of copy-pasted SQL. Your agent inspects the schema before it changes anything.

Human-reviewable

Every table, policy, key and webhook your agent creates shows up in the dashboard for you to check.

No surprise bills

Hard caps are enforced server-side. Hitting a limit blocks the request — it never creates an overage charge.

Simple, capacity-basedpricing

Every plan gets every feature. Plans only change capacity.

Start here

Free

$0forever

No credit card required
  • 1 project with its own database
  • 50 MB database storage
  • 25k API requests / mo
  • Auth, RLS, webhooks, MCP, OpenAPI
  • EU residency & one-click export
Start free
Paid

Starter

$19/ mo

Billed monthly
  • 10 projects, each with a database
  • 500 MB database storage
  • 500k API requests / mo
  • 300k writes · 20M reads / mo
  • Every feature, no add-ons
Get Starter
Most popular

Pro

$49/ mo

Billed monthly
  • 25 projects, each with a database
  • 2 GB database storage
  • 2M API requests / mo
  • 1M writes · 100M reads / mo
  • Every feature, no add-ons
Get Pro
Paid

Max

$99/ mo

Billed monthly
  • 100 projects, each with a database
  • 5 GB database storage
  • 5M API requests / mo
  • 3M writes · 300M reads / mo
  • Every feature, no add-ons
Get Max

Hard limits are enforced server-side — hitting a cap never creates an overage bill. Cancel anytime and keep your data on the free plan.

Good questionsstraight answers

Browse the field guide
What does MCPBackend actually provide?

Each project gets a SQLite-backed database with generated REST endpoints. You can configure tables, indexes, email/password auth, public or owner-based row policies, scoped machine keys and webhooks. Your coding agent manages supported setup through the remote MCP server; you inspect the result in the dashboard.

Which coding agents can connect?

Any client that supports remote Streamable HTTP MCP and the required OAuth flow. The setup guide includes configuration examples. The Agent Skill adds workflow instructions; it does not replace the authenticated MCP connection.

Is an API key safe to put in my frontend?

No. Machine API keys bypass row-level policies and use their own table permissions, so keep them server-side. Browser apps should use end-user tokens for private data, or anonymous requests only where the table policy intentionally allows public access.

What happens when I reach a usage cap?

Write and storage caps block writes; the API-request cap blocks data-API requests, including reads. The response names the quota. There are no automatic overage charges.

Can I run the exported backend myself?

Yes. Eject provides a SQL export and a compatible Node.js/SQLite runtime with project configuration. You set new secrets and users sign in again. The export does not include the hosted dashboard, MCP server, webhooks, metering or managed backups.

Ready to give your agenta real backend?

Connect in minutes. Start with one free project and the full feature set.

Start building for freeRead the setup guide