The failure it targets is the one every team recognises: documentation drifts from the code, so both the new engineer and the coding assistant work from something that was true a year ago. Regenerating the docs from the source is what makes the answers worth acting on.
The MCP entry point for Sequa, a contextual knowledge engine that generates documentation from your codebase and keeps it updated as the code changes. This repository is the connector: you add Sequa as an MCP server in Cursor, Claude, Windsurf, VS Code, Cline or any other MCP-capable client, and your assistant answers from your own architecture, docs and tickets rather than from what it can infer.
- Documentation generated from the codebase — architecture overviews, API docs, runbooks and onboarding guides — that regenerates rather than going stale
- Knowledge unified across repositories, so an answer does not depend on which repo the assistant happened to be looking at
- Plain-language questions answered with source links, for people and for agents on the same engine
- Two ways to connect: a direct URL for clients that speak HTTP, or the `@sequa-ai/sequa-mcp` proxy launched with npx for clients that expect a command
A Sequa account and — this is the step people skip — a Project created inside the Sequa app. Without the project the proxy launches but every call comes back as an authentication or project error. Once the project exists, its MCP Setup URLs page gives you the endpoint for the transport your client supports; copy that, not the example URL. Clients that take a URL get it directly; clients that want a command run `npx -y @sequa-ai/sequa-mcp@latest` with the URL as an argument.
One command — npx -y @sequa-ai/sequa-mcp@latest https://mcp.sequa.ai/v1/setup-code-assistant