Your agent builds the MCP. We run it.
Add one URL to Claude Code, Codex, Claude or Cursor and ask for the tools you want. Your own agent writes and tests them; RapidMCP hosts them on an OAuth-secured MCP server. First 1,000 tool calls free, no card.
Works with Claude Code, Codex, Claude, Cursor, VS Code and any MCP host
How it works
Add one URL to your agent
Sign up and get your MCP URL straight away. One command adds it to Claude Code or Codex, or paste it into Claude, Cursor or VS Code.
Ask your agent to build
“Build me tools that search my GitHub issues.” Your agent writes each tool as code, tests it in our sandbox and publishes it.
Use it right away
New tools are live in the same conversation. Share the MCP with your team; everyone signs in with RapidMCP.
Features
People don't want to host an MCP. They want one that doesn't exist yet. RapidMCP builds it for you and keeps it running.
Your agent is the builder
Claude Code, Codex, Claude or Cursor writes the TypeScript tools through RapidMCP's builder tools, runs them in our sandbox and fixes them until they pass.
Every tool is code
No connectors to configure. Each tool is a small script, and an HTTP call is just fetch. Click “See code” on any tool.
Secrets done right
Add API keys as env vars. Agents see the names, never the values. They're encrypted and injected only when a tool runs.
OAuth, no API keys
Every MCP is OAuth-protected. Your host asks you to sign in with RapidMCP. No headers to copy, and you can revoke access any time.
Live in the same chat
A tool your agent saves is published at once and callable seconds later, in the same conversation. Every change is versioned.
Hosted and observable
We host it. Every tool call is logged live, and every publish is versioned so you can roll back.
Built for teams
Projects have roles. Editors build, everyone else just uses the tools. You can also open an MCP to any signed-in RapidMCP user.
1,000 tool calls free
No card to start. Bring your own AI agent and pay only for the MCP layer: $0.25 per 1,000 tool calls after that.
Questions
What is RapidMCP?
RapidMCP is the MCP layer for your own AI agent. You add one URL to Claude Code, Codex, Claude, Cursor or VS Code and ask for the tools you want; your agent writes them as small TypeScript scripts through RapidMCP's builder tools, tests them in a sandbox, and RapidMCP hosts them on an OAuth-secured MCP server.
What is an MCP server?
MCP (Model Context Protocol) is an open standard that lets AI assistants call external tools. An MCP server exposes those tools, for example "search my GitHub issues" or "create a Linear ticket", so an assistant like Claude can use them during a conversation.
How do I build a custom MCP server without writing code?
Sign up to RapidMCP and you get an MCP URL straight away. Add it to your agent (for example: claude mcp add --transport http rapidmcp <url>), sign in, and ask it to build the tools you want. Your agent writes and tests each tool and it is published at once. If a tool needs an API key, you add it as an encrypted secret in RapidMCP.
How do I add a RapidMCP server to Claude?
In Claude, open Settings → Connectors → Add custom connector and paste your RapidMCP server URL. Claude will ask you to sign in with RapidMCP. In Claude Code, run: claude mcp add --transport http <name> <url>.
Does it work with Cursor and VS Code?
Yes. Any host that supports remote MCP servers over Streamable HTTP with OAuth works, including Cursor, VS Code, Claude Desktop, claude.ai and Claude Code.
How are API keys and secrets handled?
API keys are stored as env vars on your server, encrypted at rest with AES-256-GCM, and injected into a tool only when it runs. Agents building tools see the names of your env vars but never their values, and people who use your MCP never see them either.
Build your first MCP in minutes
Sign up, add your MCP to your agent with one command, and ask for the tools you want. Your first 1,000 tool calls are free, no card needed.