Kivgraph: An Open-Source Semantic Code Graph for LLM Coding Agents
ImpressiveSecret9512 · reddit · 2026-08-31
The author open-sourced Kivgraph, a local Apache-2.0 MCP server designed to solve context waste issues for LLM coding agents working on large codebases.
Core Features:
- Indexes multiple repositories into a semantic code graph.
- Supports queries like: Who calls this symbol? What breaks if I change it?
Technical Highlights:
- Edge resolution relies on language tooling (Go, TypeScript, Rust, Dart, Python) rather than simple name matching.
Benchmark Results (37 repos, 29 questions):
- Kivgraph: 28/29 exact answers, 35,961 tokens.
- grep + reading: 28/29 exact answers, 267,980 tokens.
- Kivgraph reduced token usage by 7.4x overall. The project runs entirely locally without an API key.
More from coding & agent
- 818 Open-Source Cybersecurity Skills Empower AI Agents Across 6 Frameworks — tom_doerr · 2026-08-31
- Using Grok bot to handle web scraping tasks without APIs — PTrubey · 2026-08-31
- Why run multiple coding sessions with AI agents? — Jason-Ping · 2026-08-31
- Doug and Hugo Launch Enterprise AI Agent Building Course — hugobowne · 2026-08-31
- From RAG to Agents: Rearranging Search Building Blocks for Better Retrieval — hugobowne · 2026-08-31
- Open source Gambit project uses cross-domain frameworks to aid strategic planning — kidwonder · 2026-08-31