MCP Server
A Model Context Protocol server that connects agents to your tools and data. Deploy it to your own cloud and point your client at the URL.
About this product
Model Context Protocol is the standard way to give an agent access to tools and data it does not have built in. An MCP server exposes your systems as typed tools; the agent discovers them and calls them. This is the server we built and run ourselves, packaged so you can deploy your own.
It ships as a TypeScript codebase with its Dockerfile and a parameterised Cloud Build config. The transport is streamable HTTP and the server is stateless, so it scales to zero and back without session affinity. Deploy it to Cloud Run — or any host that runs a container — point your client at the URL, and the tools appear.
The parts that are tedious to get right are already done: tool schemas written with full parameter descriptions, bearer-token authentication on every request, per-request credential passing so one deployment can serve more than one backend, and a health endpoint you can point a monitor at.
What's included
- TypeScript source for the server, with its Dockerfile and a parameterised Cloud Build config.
- Streamable HTTP transport, stateless by design, so the service scales to zero between requests.
- Tool definitions covering n8n workflow management and Google Cloud infrastructure.
- Bearer-token authentication and per-request credential headers, so one deployment can serve more than one backend.
- A deploy guide for Cloud Run, including the scale-to-zero settings we run in production.
- A health endpoint, and client registration snippets for the Claude Code CLI.
Compatibility & requirements
- LanguageTypeScript, built on Node.js 20
- TransportStreamable HTTP, per the MCP specification
- Deploy targetAny container host — Cloud Run documented
- BuildDocker, or a managed container build service
- ClientsClaude Code CLI, or any MCP client that speaks streamable HTTP
- DeliveryZIP archive, download link by email