Graft, a new tool from Nanonets, aims to significantly improve the efficiency and contextual understanding of coding AI agents like Claude Code, Cursor, and Gemini. By building a persistent, explorable graph of a codebase, Graft reduces the need for agents to re-discover repository information with each task. This approach has demonstrated a 42% reduction in token usage and improved SWE-bench performance, resolving 66% of instances compared to Claude Code's 54%. The tool integrates directly with agents and stores the codebase graph as linked markdown files within the repository, allowing for easy sharing and version control via Git. AI
IMPACT Enhances coding AI agents by reducing token usage and improving contextual understanding, potentially accelerating development workflows.
RANK_REASON This is a new software tool release that enhances existing AI models.
Read on HN — claude-code stories →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
