LibreAI / Dev Tools / Graft
// devtools · trailhq/Graft

Graft

A codebase knowledge graph that makes Claude Code, Cursor, Codex and Gemini faster, cheaper and repo-aware.

★ 9.8k stars⑂ 885 forksMITupdated 2026-10
▦ visual — this project's README ships no embeddable screenshot,
so here's a placeholder. The guide below is complete regardless.
// project facts

Project facts

Graft is an open-source dev tools project by trailhq (trailhq/Graft on GitHub). It has 9,787 GitHub stars and 885 forks, is released under the MIT license, and was last updated on GitHub on 8 October 2026.

Repository
trailhq/Graft
GitHub stars
9,787
Forks
885
License
MIT
Last updated on GitHub
8 October 2026
Category
Dev Tools
Listing
Full install guide
Official site
Official site
// how it works

How it works

  1. Clarify your goal

    Pick the coding agent you use and the LLM provider whose key will write Graft's optional deep summaries; the structural graph itself needs no key.

  2. Get the project

    Clone trailhq/Graft from GitHub (or use the install path in the run steps below) and skim its README for the parts relevant to your goal.

  3. Install and configure

    Install the CLI globally from npm, then run graft init in your repo to drop in the wiring for your agent.

  4. Run your first task

    Run graft build to generate the graph, then ask your agent a codebase question and let it follow the graph instead of grepping from scratch.

// features

Why people love it

Plain-English code graph

Each node explains in plain English what part of the system does and how it connects, the way a senior engineer would — not just a dump of symbols.

Always-fresh graph

Every query rebuilds the graph against the working tree first (about 3ms when nothing moved), so answers include uncommitted edits with no stale index.

Local cache, tiny footprint

graft build writes a regenerable local graft/ cache (gitignored, like node_modules); you only commit the small init wiring — no database or server.

Bring your own provider

Summaries are written by any provider you choose — OpenAI, Anthropic natively, OpenRouter, Fireworks, Groq, OrcaRouter, a LiteLLM proxy or a local model — under your own key.

Deterministic structural graph

The core graph, graft check and graft ask are deterministic tree-sitter and never call a model, so they work offline with no key.

Deep mode and LSP edges

graft build --deep adds LLM concept nodes and per-symbol summaries; --lsp adds opt-in compiler-grade edges on top of scope-aware cross-file extraction.

Advertisement
// run it

Run it yourself

  1. Clone the repository

    Get the source from the official GitHub repository for Graft.

    git clone https://github.com/trailhq/Graft.git
    cd Graft
  2. Install the CLI

    Install Graft globally from npm, exactly as the README's quick start shows (or use npx without installing).

    npm install -g @nanonets/graft
  3. Initialise and build

    Wire Graft into your repo for your coding agent, then build the codebase graph.

    npx @nanonets/graft init
    graft build
// use cases

What you can build

Advertisement
// faq

Questions, answered

Who maintains Graft?

Trailhq, through the trailhq/Graft repository, with the product site at trailhq.com/graft.

Is Graft free to use?

Yes — the project is open source under the MIT license as recorded for this listing. You can use, modify and distribute it under the MIT terms.

Does Graft need a paid API key?

The structural graph needs no API key at all. Only the optional --deep summaries call an LLM, using your own provider key, so any cost is your provider usage.

How does Graft compare to a vector-embedding code index?

Embedding indexes need similarity search and a warm index that can go stale. Graft is a real, readable graph of linked files your agent opens and follows, rebuilt deterministically on every query — no embeddings and no server.

How do I get started with Graft?

Clone the repository (or use the documented installer), follow the install and configure steps on this page, and run the smallest real task first — run graft build to generate the graph, then ask your agent a codebase question and let it follow the graph instead of grepping from scratch.

// keep exploring

Similar tools

Dev Tools

cc-connect

Bridge local AI coding agents (Claude Code, Cursor, Gemini CLI, Codex) to messaging platforms (Feishu/Lark, DingTalk, Slack, Telegram, Discord, LINE, WeChat …

★ 15.7k starsUnknown
View guide →
Dev Tools

cocoindex-code

Lightweight AST-based semantic code search CLI (ccc) that makes coding agents faster and cheaper

★ 2.8k starsApache-2.0
View guide →
Dev Tools

gentle-ai

Gentle-AI configures the AI coding agents you already use: Claude Code, Cursor, OpenCode, Codex, Pi, and more. Choose persistent memory, Organic-Driven Devel…

★ 7.5k starsMIT
View guide →
Dev Tools

ClaudeBar

Menu-bar tracker for your AI coding quotas — Claude, Codex, Gemini and more.

★ 1.5k starsCustom
View guide →
Dev Tools

claude-octopus

Claude Octopus orchestrates Claude, Codex, Copilot, Gemini, and Ollama on one task — and surfaces disagreement

★ 4.2k starsMIT
View guide →
Dev Tools

avibe

A local-first Agent OS — drive Claude Code, Codex and OpenCode from your browser.

★ 606 starsMIT
View guide →
// alternatives

Alternatives in Dev Tools

The most-starred other Dev Tools projects in the LibreAI catalogue — the realistic shortlist if Graft is not the right fit.

Get Graft

Free and open — grab the source, star the repo, and start creating.