Ship your AI the docs —
and the graph.
One command downloads a library's docs as Markdown plus a pre-built knowledge graph, pinned to a release. Far better context for your AI, at a fraction of the tokens.
$ uvx graflet react ✔ Resolved react@19.1.0 (pinned) ✔ Downloaded docs → ./graflet/react/*.md (312 files) ✔ Fetched knowledge graph → graph.json (1,204 nodes · 3,880 edges) ✔ Aligned to release react@19.1.0 Done in 8m 12s · saved ~$0.42 in build cost▋
Knowledge-graph catalog
aligned to pinned releases| Library | Version | GraphScore | Tokens saved | Graph size | Updated | Command |
|---|---|---|---|---|---|---|
| Loading catalog… | ||||||
Docs as Markdown
Every page of the official docs, downloaded as clean Markdown you can grep, feed, and version — pinned to the exact release you install.
Knowledge graph
A pre-built graph of concepts, APIs, and how they connect — as an interactive view and a machine-readable graph.json. This is the context your AI actually needs.
Live from the catalog · — until the first graph is measured
How it works
Run one command
uvx graflet react — no config, no account to start.
Sign in with GitHub
One click, and only to download the graph. Install, browse and copy stay free.
We fetch both
Docs pulled from the upstream project, the knowledge graph from our backend.
Land locally, aligned
docs/*.md and graph.json on disk, pinned to the same release.
Raw docs are flat. A graph knows how things connect.
The graph encodes relationships — which API calls which, what depends on what, which concepts connect — so your AI retrieves the right context instead of dumping whole pages. Cheaper prompts, sharper answers, fewer hallucinations.
Support the project
Free and open source. No paid plans — ever.