Describe your tool.
We host the server.
AI uses it.
From a prompt to a production MCP server — OAuth 2.1, Streamable HTTP, ready for Claude, Cursor and ChatGPT.
Build an MCP server that
searches our Notion workspace.
Tools: search_pages, get_page
Needs: NOTION_API_KEY## how_it_works
Three steps. No JSON to write, no Docker to manage.
Describe your tool
A sentence is enough. List your secrets and which APIs to call.
We generate, check, deploy
Claude writes the spec. We render TypeScript, run static checks, build a container, deploy to your subdomain.
Install in your client
Copy the snippet into Claude Desktop, Cursor or ChatGPT. OAuth flow on first use.
## verify_it_yourself
Don't take our word for it.
Every claim links to its proof.
OAuth 2.1 authorization server
PKCE, Dynamic Client Registration and Resource Indicators (RFC 8707) in front of every server.
Read the auth docs →Live system status
Uptime and incident history, public. If a build queue slows down, you see it before we tell you.
Check status now →Per-server container isolation
Every generated server runs in its own Docker container. Secrets AES-256-GCM encrypted at rest.
Security architecture →Works with Claude Desktop, Cursor, ChatGPT, VS Code Copilot and Continue.dev — anything that speaks MCP.
## use_cases
Wrap any HTTP API.
From one prompt.
Each shipped from a single prompt.
PostgreSQL
Read-only access to your tables with schema introspection.
Salesforce
Query opportunities, accounts and leads from Claude.
Notion
Search pages, read content, append blocks.
GitHub
List issues, search code, post comments. Scoped to one repo.
Stripe
Look up charges, customers, refunds (read-only by default).
Custom REST
Wrap any HTTP API behind one prompt-defined tool surface.
## marketplace
Skip the prompt.
Fork what works.
The marketplace is a library of working MCP servers. Fork one, paste your credentials, deploy. Or publish yours and let others build on it.
Fork and own
Start from a server someone already shipped. Fork it, paste your own credentials, deploy. No prompt required.
Secrets never travel
A template carries the spec and generated code, never the author's API keys. You add your own on fork.
Open from day one
Publish a server you built and anyone can fork it. The marketplace is young — early templates set the standard.
## pricing
Pay for tool calls.
Not for boilerplate.
For trying things out and shipping single-user tools.
- 1 MCP server
- 100,000 tool calls / month
- 5 prompt analyses / day
- BuildMyMCP subdomain
- Community support
For solo founders and small teams shipping production tools.
- 5 MCP servers
- 1M tool calls / month
- 40 prompt analyses / day
- Priority build queue
- Custom domain · coming soon
- Email support, 1 business-day response
## faq
Questions, answered straight.
What is MCP?
Model Context Protocol — an open standard from Anthropic for connecting AI assistants to external tools, data and APIs over a transport like Streamable HTTP.
Do I need to write code?
No. You describe the tool in natural language. We generate the TypeScript server, run static checks, build a Docker image and deploy it to a public OAuth-protected URL.
Which clients work?
Claude Desktop, Cursor, ChatGPT Custom Connectors, VS Code Copilot, Continue.dev — anything that speaks the MCP spec.
How is auth handled?
Every generated server is an OAuth 2.1 Resource Server. Our control plane is the Authorization Server (PKCE + Dynamic Client Registration + Resource Indicators per RFC 8707).
Can I self-host?
Yes. The runner is a plain Docker container; the control plane is open to BYO Postgres + Redis. See the self-hosting guide in docs.
What about secrets?
AES-256-GCM at rest in Postgres, injected as environment variables into the runtime container. Never logged, never echoed back.
Cold starts?
No cold starts — each server runs in a container that stays warm (auto-restarts on failure), so there is no spin-up delay on a tool call.
Rate limits?
Every request is gated by OAuth 2.1 before it reaches your container. Configurable per-server rate limiting is on the roadmap.
How fast is generation?
Spec to image to live URL typically completes in 45-90 seconds.
Logs and metrics?
Live build-log streaming to the dashboard, plus structured tool-call metrics — latency, error rate and per-tool throughput — viewable per server.
What if I cancel?
You can export the full TypeScript source of every server you built. No vendor lock-in.
Custom domain?
On the roadmap. Today every server is reachable at a buildmymcpserver.com URL with TLS; bring-your-own-domain support is coming.
Your first server is one prompt away.
Free tier, full source export, no lock-in. If it doesn't work for you, take the TypeScript and leave.