#363 · Primary category: AI Coding Assistants
GrapeRoot
Compounding Context for AI Coding Assistants — MCP graph engine for Claude Code, Cursor, Copilot, Gemini, OpenCode
Project last updated:08/04/26
GitHub Stars
1.0K
Forks
123
Contributors
4
License
Apache-2.0
Why we included this project
If your coding assistant spends its first several turns hunting for the right files, GrapeRoot is worth a look. It sits between you and the assistant, building a semantic graph of the project's files, symbols, imports, and call chains, then packing the relevant code into each prompt before the assistant ever sees your question. Because the graph remembers which files were already read and edited, context builds up across a session instead of being rebuilt from scratch, which cuts turns, tokens, and cost while keeping answer quality steady. It runs fully locally and plugs into a wide range of assistants, from Claude Code and Cursor to Copilot and Gemini, so teams can adopt it without abandoning their current workflow. For anyone working in a large or unfamiliar codebase, that's a practical way to make the assistant start reasoning immediately.
Articles for this project
No articles for this project yet.
To suggest a topic or contribute an article, contact us.
Related projects in this category
superpowers
An agentic skills framework & software development methodology that works.
ECC
The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.
skills
Skills for Real Engineers. Straight from my .agents directory.
opencode
The open source coding agent.
andrej-karpathy-skills
A single CLAUDE.md file to improve Claude Code behavior, derived from Andrej Karpathy's observations on LLM coding pitfalls.