Star Documentation
HTTP Endpoints
Expose your own HTTP APIs to the assistant as named tools — no integration required.
HTTP Endpoints let you turn your own APIs into tools. Instead of waiting for a built-in integration, you describe an HTTP request once and it becomes a named tool the assistant can call — perfect for internal services, webhooks, or any third-party API with an HTTP interface.
How it works
Each endpoint you add registers as its own tool. You define:
- the method and URL to call (with placeholders the model fills in at call time);
- any headers — including auth like an API key or bearer token, stored securely;
- the inputs the tool accepts, which become the arguments the model provides;
- how the request body is built from those inputs.
When the model calls the tool, Ganju makes the HTTP request on your behalf and returns the response for the model to use. Requests are screened against SSRF (they can’t be pointed at internal addresses) and rate-limited to protect your services.
When to use it
Reach for an HTTP endpoint when you want the assistant to hit an API we don’t ship as a first-class integration — your own backend, a CRM, an internal lookup, or a partner API. For a vendor that publishes an official remote MCP server, prefer connecting that instead (see GitHub and Notion), since it brings a whole maintained toolset rather than a single request.
Last updated