ByteZone
Back to listClaim this project

NanoNetsDeveloper tools

Graft

Open-source context layer that builds structural codebase maps for AI coding agents

  • CLI
  • Library/framework
  • Windows
  • macOS
  • Linux
  • Self-hostable
  • Runs locally
  • Docker supported
  • Needs API key
Open websiteView on GitHub
Graft screenshot
Popularity
8.5k Stars
GitHub stars
Recent activity
9/18/2026
Updated in the last 30 days
License
MIT
Permissive

Why it matters

We look beyond stars: what problem it solves, whether it creates real utility, and what makes its approach worth noticing.

Last 90 days

Problem

Coding agents re-explore the codebase from scratch for every task, incurring high overhead in tokens, tool calls, and latency.

Practical value

Generates a local Markdown context graph via tree-sitter and LLMs, reducing tool calls by 46% and token usage by 42% in benchmarks.

Innovation / differentiation

Replaces complex vector databases with plain Markdown files stored directly in the repository and synchronized via Git.

Leverage potential

Distributed as a Node.js CLI with one-command initialization, integrating via hooks and MCP servers without background daemons.

Why now

As terminal-based coding agents rapidly proliferate, context efficiency and exploration overhead in large codebases have become key pain points.

Community activity

In the last 90 days there were 39 new issues and 98 pull requests; the bounded issue/PR samples include 27 issue authors and 21 PR contributors, with 0 releases.

Maintainer responsiveness

The 39-issue window sample had a 54% close rate, and the 98-pull-request sample had a 45% merge rate. Maintainer-response observations covered 100% of that issue sample, with a 36% response rate and median first response of 3.6 days.

21 contributors in PR sample0 releasesIssue response rate 36% · sample 39 (100% coverage)Median first response 3.6dPR merge rate 45% · 98 sampled in window

Key highlights

  • Markdown context graph: Converts codebase into interconnected markdown documentation nodes
  • Deterministic parsing: Uses Tree-sitter to build structural code graphs without LLM overhead
  • Agent integration: Integrates directly with Claude Code, Cursor, Gemini, and MCP servers

Quick start

How it is installed, how hard it is, and where to start.

Where it runs

To be confirmed

Difficulty

To be confirmed

Docker supportedSelf-hostableRuns locally
  1. 01Install the Graft package globally using npm: npm install -g @nanonets/graft
  2. 02Run the initialization command in your repository root to build the graph and configure agents: graft init
  3. 03Commit the generated configuration and hooks to Git: git add .claude && git commit -m "wire in graft"

Best for

  • Teams that want data on their own servers
  • Developers who want to try it on their machine
  • People who prefer Docker deploys

More about it

Graft serves as a context understanding layer for AI coding agents. Instead of forcing AI tools to re-explore files on every task, Graft analyzes the codebase once and builds a network of markdown documentation files detailing systems and API relations. Agents can directly reference these files to execute actions faster and use fewer tokens while maintaining code correctness.

Sources

Each field shows its status and source — expand to review.

14 · Expand

  • capability tags

    Verified

    Source: admin_cms · cms editor · 8/30/2026

  • editor note

    Verified

    {"en":"Graft solves the issue of AI coding agents re-exploring codebases on every prompt. By using Tree-sitter for AST parsing and custom LLM calls for summaries, it provides agents with structured context to skip unnecessary file searches.","zh":"Graft 解决了 AI 编程工具每次执行任务都需要重新探索代码库的问题。它使用 Tree-sitter 进行确定性语法解析，并利用自定义 LLM 生成模块说明文档，减少模型盲目搜索导致的消耗。"}

    Source: admin_cms · cms editor · 8/30/2026

  • how to use

    Verified

    {"steps":[{"en":"Install the Graft package globally using npm: `npm install -g @nanonets/graft`","zh":"使用 npm 全局安装 Graft 工具包：`npm install -g @nanonets/graft`"},{"en":"Run the initialization command in your repository root to build the graph and configure agents: `graft init`","zh":"在项目根目录下运行初始化命令，生成图表并配置 AI 助手：`graft init`"},{"en":"Commit the generated configuration and hooks to Git: `git add .claude && git commit -m \"wire in graft\"`","zh":"将生成的配置文件与钩子提交至 Git 仓库：`git add .claude && git commit -m \"wire in graft\"`"}],"installAt":"unknown","difficulty":"unknown"}

    Source: admin_cms · cms editor · 8/30/2026

  • intro

    Verified

    {"en":"Graft serves as a context understanding layer for AI coding agents. Instead of forcing AI tools to re-explore files on every task, Graft analyzes the codebase once and builds a network of markdown documentation files detailing systems and API relations. Agents can directly reference these files to execute actions faster and use fewer tokens while maintaining code correctness.","zh":"Graft 是一个面向 AI 编程 Agent 的上下文映射工具。在开发过程中，AI 助手往往在处理新任务时需要频繁寻找文件和调用工具。Graft 能够一次性分析代码库并生成包含架构与概念说明的 Markdown 关联网络。Agent 可以直接查阅该文件目录，无需重复执行全局搜索，从而在保持准确度的同时缩短响应时间并降低 Token 消耗。"}

    Source: admin_cms · cms editor · 8/30/2026

  • License

    Verified

    MIT

    Source: GitHub API · license.spdx_id=MIT · 9/18/2026

  • needs api key

    Verified

    Yes

    Source: admin_cms · cms editor · 8/30/2026

  • One-liner

    Verified

    {"en":"Open-source context layer that builds structural codebase maps for AI coding agents","zh":"为 AI 编程 Agent 提供代码库结构与上下文映射的开源工具"}

    Source: admin_cms · cms editor · 8/30/2026

  • Platforms

    Verified

    windows, macos, linux

    Source: admin_cms · cms editor · 8/30/2026

  • Category hint

    Inferred from materials

    ai-apps

    Source: Project README · hint=ai-apps · 8/17/2026

  • product forms

    Verified

    cli, library_framework

    Source: admin_cms · cms editor · 8/30/2026

  • role tags

    Verified

    Source: admin_cms · cms editor · 8/30/2026

  • supports docker

    Verified

    Yes

    Source: Repository file · dockerfile=true; compose=false · 9/18/2026

  • supports local

    Verified

    Yes

    Source: admin_cms · cms editor · 8/30/2026

  • supports self host

    Verified

    Yes

    Source: admin_cms · cms editor · 8/30/2026

Related projects

Other verified projects matched by category, capabilities, and intended roles.

claude-code

An agentic coding tool for the terminal that understands codebases, handles routine development tasks, explains code, and works with Git workflows.

Same Developer tools categoryActive in the last 30 days

spec-kit

An open-source CLI toolkit for Spec-Driven Development that guides AI coding agents through structured specification, planning, and code generation steps.

Same Developer tools categoryActive in the last 30 days

llama.cpp

A lightweight LLM inference engine implemented in C/C++, supporting various hardware backends and model quantization.

Same Developer tools categoryActive in the last 30 days

codex

OpenAI’s coding agent that runs locally, with terminal, IDE, and desktop-app entry points.

Same Developer tools categoryActive in the last 30 days

graphify

Convert codebases, docs, and configs into a queryable knowledge graph for AI coding assistants

Same Developer tools categoryActive in the last 30 days

supabase

An open-source Postgres development platform offering auto-generated APIs, authentication, and real-time database subscriptions.

Same Developer tools categoryActive in the last 30 days