Back to portfolio

Case study · Bridge Network

codegraph

A pre-indexed code knowledge graph for local AI coding agents — fewer tokens, fewer tool calls.

Role
Builder / maintainer
Building
Knowledge Graph Spans

Problem

Coding agents burn tokens and round-trips rediscovering repo structure. A local, pre-indexed graph can answer structural questions without shipping the whole codebase into context.

Daniel's role

Designed and built a knowledge-graph index that works with Claude Code, Codex, Cursor, and OpenCode — running 100% locally.

System diagram

codegraph local knowledge graphLocal index · fewer tokens · fewer tool calls

Stack

  • TypeScript / Node
  • Local indexing
  • Knowledge graphs
  • Agent tooling
  • CLI

Outcome

Open-source local graph that reduces redundant tool calls for agent workflows. See the public repo for current status.

Local-first RAG for code: indexing strategy and retrieval quality beat throwing more context at the model.

Links

Field notes (measurements)

  • Runtime

    100% local

  • Agents

    4

Find this building on the city plan →