Generate an mcp.json discovery file that describes your public MCP server endpoint, so AI clients and agents can find and connect to it.
How to use the mcp.json generator
- Enter your server name, endpoint URL and auth method.
- Copy the generated file.
- Serve it from your domain's well-known path.
What MCP is
The Model Context Protocol lets AI assistants call tools and read data from external servers. A discovery file tells clients where your server is and how to connect.
What each field means
name is a short machine id, lowercase with no spaces. title is the display name people see. description should say in one line what the tools let an assistant do. remotes lists the endpoint: its transport (Streamable HTTP or SSE), its URL and the auth type (none, OAuth or a bearer token). websiteUrl points to your docs, and contact is where developers reach you. The $schema line names the server schema the file follows, so tools can validate it. Serve the result at /.well-known/mcp.json with a JSON content type.
Getting it right
Use Streamable HTTP for new servers; the MCP spec replaced the older SSE transport with it in 2025, though many clients still accept SSE. The endpoint must be HTTPS and reachable from the internet, and the URL here must be the actual MCP path, not your home page. If you choose OAuth, make sure your server publishes its OAuth metadata, because clients will look for it. Discovery conventions are still settling, so also list your server in the public MCP registry and link it from your docs.