ChubThe all-in-one infrastructure layer for AI-assisted development

The All-in-One Agent Layer ​

Most tools do one thing: serve docs, or track usage, or scan secrets, or sync configs. Chub does all four — because context, tracking, security, and team knowledge are one problem, not four.

Install ​

npm install -g @nrl-ai/chub

pip install chub

cargo install chub

brew install nrl-ai/tap/chub

Or download a prebuilt binary from GitHub Releases — single 10 MB binary, no runtime dependencies.

Quick Start ​

1. Give your agents accurate docs ​

chub search "stripe payments"              # find docs
chub get openai/chat --lang python         # fetch a doc
chub mcp                                   # start MCP server for agents

2. Track what your agents do ​

chub track enable                          # install hooks (auto-detects agent)
# ... use your AI agent as normal ...
chub track status                          # see active session
chub track report                          # costs, tokens, models, tools
chub track dashboard                       # web dashboard at localhost:4243

3. Catch leaked secrets ​

chub scan secrets git                          # scan git history
chub scan secrets git --staged                 # pre-commit hook mode
chub scan secrets dir ./src                    # scan a directory

4. Build team knowledge ​

chub init --from-deps                      # create .chub/, auto-pin docs
chub annotate openai/chat "Use streaming"  # team annotation
  --kind practice --team
chub agent-config sync                     # generate CLAUDE.md, .cursorrules

MCP Setup ​

Add to your MCP config and your AI agent gets instant access to context, tracking, and team knowledge:

{
  "mcpServers": {
    "chub": {
      "command": "chub",
      "args": ["mcp"]
    }
  }
}

Works with Claude Code, Cursor, Windsurf, Copilot, Gemini CLI, Kiro, Codex, Cline, Roo Code, Augment, Continue.dev, and Aider. See Agent Integrations for setup guides.

8 MCP tools: chub_search, chub_get, chub_list, chub_context, chub_pins, chub_annotate, chub_feedback, chub_track

Benchmarks ​

Context vs Context Hub — 1,553 docs, median of 5 runs ​

Secret scanning — 10 real public repos, median of 3 runs ​

Directory scan (chub vs gitleaks v8.30.1):

Chub is 2–4x faster on repos up to ~7k files. gitleaks leads on very large monorepos. See full table.

What makes Chub different ​

Built on Context Hub by Andrew Ng — fully format-compatible. Complementary to Context7. Read the full story.