// your docs as a directory your agent can read
Your docs,
agent ready.
Distill your docs into a tree of markdown files your coding agent can intuitively cat and grep, plus a ⌘K overlay that searches your docs site. The core is framework-neutral, the Astro integration is first-class, and you never pay by the token.
- build the digest
// offline — distils your markdown into the .hev-ask/ tree ask digest build - add the overlay
// any layout — the Astro integration, or one script elsewhere <SearchOverlay /> - optional — single-turn Q&A
// add a key to stream a grounded answer on Enter ANTHROPIC_API_KEY=sk-…
That's the whole setup, the same on any framework — this site runs on it. Keyword search needs no API key; add one only for single-turn Q&A. Quick start →
// and for their agents
- install the CLI — no config
go install github.com/hev/ask/cmd/ask@latest - browse any hev-ask site as a directory
ask --endpoint https://yourdocs.com/api/ask tree ask --endpoint https://yourdocs.com/api/ask cat overview/quick-start
Your docs become a directory any agent can navigate: tree the map, cat a section, grep for a string — every read keyless, every citation a CI-verified anchor. Or wire up MCP: one tool hydrates the whole tree to disk and the agent reads it with the file tools it already has. CLI reference →
- readers ⌘K — instant keyword search; Enter streams a grounded Claude answer
- agents ask tree · cat · grep — your docs as a directory, keyless, from any shell
- cited every answer deep-links to the exact heading — anchors verified in CI
// the pitch, second
ask is free and open source, built by hev mind — Adam Hevenor's search engineering practice. If you want to go deeper than your docs — training your team on agent-ready design, hands-on search engineering, or taking a retrieval workload to production:
training — make your product agent-ready work with the practice production retrieval — hev layer