Getting Started

Canopy is a local code-intelligence engine. It indexes your repository — AST symbols, the dependency graph, full-text and semantic search — and answers questions about it through one MCP tool, canopy_answer. Everything runs on your machine.

The whole path is three commands:

# 1. Install (see the install guide for your platform)
curl -fsSL https://downloads.canopy.gulfshieldtech.com/releases/latest/canopy-x86_64-unknown-linux-gnu.tar.gz \
  | tar -xz -C ~/.local/bin

# 2. Set up — from inside the repo you want to index
cd /path/to/your/repo
canopy setup

# 3. Ask
canopy answer "what depends on the payment module?"

canopy setup is the whole onboarding in one command: it walks license, embedding runtime, summary backend, cache, editor wiring, and indexing — each step with a one-Enter default. When it finishes, your repo is indexed and your AI editor is already connected. canopy setup --yes accepts every default unattended.

Check readiness any time:

canopy doctor

It reports your license tier, index status, the active summary backend, and exactly what (if anything) leaves your machine.

The guides

1. Install Canopy — binaries for Linux, macOS, and Windows; verify with canopy --version.
2. Index Your First Repo — what canopy setup and canopy index actually build, and the defaults.
3. Connect to Claude Code — setup wires it automatically; the manual recipe if you prefer.
4. Your First Questions — the question shapes that get precise answers, and why phrasing matters.

What Canopy is (and isn’t)

Canopy does not include an AI model. It is a context engine — it gives your existing AI tools precise, structured facts about your codebase: what depends on a file, where a symbol is defined, what tests cover it, whether it’s dead code. Answers come with file:line citations. And when nothing relevant exists, Canopy says “No strong match in this repository” instead of guessing — an empty answer is a real signal, not noise.

For the full story, see Concepts: What Canopy Is and Isn’t.