Limited time offer: 50% off your first 3 months
Deploy hosted MCP servers from your OpenAPI spec
Turn your existing API into a production-ready MCP server in minutes — no infra to manage, built for agents.
$
Paste your spec to configure endpoints, auth, and hosting — then sign up to publish.
No MCP runtime to ship — just your OpenAPI.
No service accounts — OAuth 2.1, real users.
No agent code in your API — keep it pure REST.
How it works
OpenAPI to production MCP, in three steps.
No infrastructure to operate and no auth code in your backend — MCP Stack runs the runtime, the gateway, and the logs.
Import your OpenAPI
Paste a URL or upload a spec. MCP Stack generates typed MCP tools you can name, scope, and add instructions to.
Add OAuth with Gateway
Put Gateway in front and agents call your API as the signed-in user — your existing scopes and audit trail, no shared service accounts.
Connect agents & clients
Publish a stable MCP URL for Claude, Cursor, and ChatGPT — or embed an agent in your own app with the SDK.
Per-user identity, built in
Agents call your API as the signed-in user.
No shared service accounts. Gateway brokers OAuth and forwards each user's own access token to your API on every tool call — so your existing scopes, RBAC, and audit trail keep working.
- OAuth 2.1 with protected-resource metadata and dynamic client registration
- Per-user grants brokered and stored server-side, never in the browser
- Static-header and public-API modes when you don't need user auth
Global edge MCP
Create once. Endpoints live near every agent.
Multi-region runtime with regional health, gateway routing, and tool-call telemetry — out of the box.
@mcpstack/agent-sdk
Embed agents in your app.
Pass your signed-in user, render <McpStackChat />, and the gateway handles the token exchange. Add client tools to drive your UI and cap hosted AI spend per user — no billing controls to build.
import { McpStackChat } from "@mcpstack/agent-sdk/react-embed";
<McpStackChat
agentId="ag_invoices"
apiKey={"pk_mcpstack_...."}
clientTools={clientTools}
auth={{
mode: "app-token",
getToken: async () => session.accessToken,
}}
/>
Production-grade by default
Everything you need to run MCP in production.
Auth, hosting, and observability come standard — so you ship the integration, not the infrastructure.
OAuth 2.1 gateway
Protected-resource metadata, dynamic client registration, and per-user token brokering for external MCP clients.
Global edge hosting
Generated runtimes publish across managed edge regions behind a stable public MCP URL with TLS.
Request & tool-call logs
Every call traced with client, tool, auth state, latency, and status — in the dashboard, CLI, or API.
Agents with budgets
Configure prompts, models, and attached tools, then cap hosted AI spend per user with monthly budgets.
Client tools & app context
Let an embedded agent drive your UI and read app state alongside server-side MCP tools.
CLI & REST API
Create servers, publish, and stream logs with the mcpstack CLI or automate everything over REST.
Simple, usage-based
Publish MCP servers with OAuth in minutes.
Every plan includes a globally published MCP server — MCP Stack runs the runtime in regions close to your agents, with a stable public MCP URL, OAuth, and request logs included.
Developer
/ month · 50% off first 3 months
$49/month
$24.50 / month · 50% off first 3 months
- 1 edge-hosted MCP server
- 10,000 tool calls included
- 30s max tool call
- $2 per 1,000 additional tool calls
Team
/ month · 50% off first 3 months
$199/month
$99.50 / month · 50% off first 3 months
- 10 edge-hosted MCP servers
- 100,000 tool calls included
- 60s max tool call
- $2 per 1,000 additional tool calls
Enterprise
invoice billing
Custom
- Unlimited hosted, self-hosted, or hybrid MCP servers
- Stripe invoices with NET terms
- White-glove onboarding and support
- Custom contract limits
New field note
Building a ChatGPT-ready SaaS app with MCP Stack.
Follow the PetalPal build from a normal signed-in product dashboard through SqlOS OAuth, FGA, OpenAPI, Gateway, CLI checks, screenshots, and the ChatGPT connection path.
Your API. Their agents. Live in under a minute.
Free for 7 days. Direct line to founders if you need it.