crabcc

Symbol index for AI coding agents.

4412× faster than grep. 85% fewer bytes sent to the LLM. MCP-native.

Quick start

# Install
curl -fsSL https://raw.githubusercontent.com/crabcc-labs/crabcc/main/install.sh | bash

# Index a repo
cd your-project
crabcc index

# Query
crabcc lookup sym ParseError        # find symbol definition
crabcc lookup refs ParseError       # find all references
crabcc lookup outline src/main.rs   # structural outline

# MCP server (for AI agent tools)
crabcc serve

Core commands

Command	What
crabcc index	Build symbol database (~5-30s for monorepo)
crabcc lookup sym <name>	Find symbol definition
crabcc lookup refs <name>	Find all references
crabcc lookup outline <file>	Structural outline
crabcc graph	Call-graph operations
crabcc memory	AI memory (past findings)
crabcc serve	MCP server + localhost viewer

Architecture

6 crates:
  crabcc-cli     — CLI binary
  crabcc-compact — memory compaction
  crabcc-core    — indexing engine (tree-sitter based)
  crabcc-fetch   — URL fetch + crawl
  crabcc-mcp     — MCP server
  crabcc-memory  — AI memory (BM25 + sqlite-vec)

Install options

brew install crabcc-labs/tap/crabcc
cargo install crabcc-cli
# or: curl -fsSL https://raw.githubusercontent.com/crabcc-labs/crabcc/main/install.sh | bash

License

MIT