What it does
Cortex is a command-line tool driven by one cortex.config.yml file. In it you list your API sources, including OpenAPI, AsyncAPI, GraphQL, Protocol Buffer and OpenRPC specifications, along with Markdown pages, and a single command generates every configured output. The outputs are static HTML documentation with interactive API reference pages, SDKs for TypeScript, Python, Go, Java, Kotlin, Ruby, PHP, C#, Rust, C++ and C, and an MCP server that bundles the specifications, SDK guides and project docs for AI agents. The generated server does not call gRPC methods; it exposes Protocol Buffer files as resources an agent can read. Day to day, cortex init scaffolds a project, cortex docs serve previews it locally with file watching, cortex docs build exports static files for any web host, and cortex publish sends generated packages to registries, with a dry-run option. For a public GitHub repository, one deploy command publishes a documentation site on a cortexdocs.dev address, and a second command connects AI clients to the current docs over local MCP. It needs Node.js 20 or later and npm 10 or later, and some generated SDKs need the usual compiler or package manager for their language. The project is released under the MIT License.
Capabilities
- One project configuration drives docs, SDKs and MCP output
- SDKs for eleven languages, including TypeScript, Python, Go and Rust
- Interactive API reference built as static HTML
- MCP server with typed tools plus embedded specs and guides
- Sparse template overrides to customize generated output
- Local preview with file watching and a publish dry run
When to use it
- Publishing API docs, client libraries and an agent-facing MCP server from the same specification files
- Giving AI clients current access to an API's specs and guides through MCP
- Hosting a free documentation site for a project in a public GitHub repository