GitSense

Build repository intelligence for AI agents.

Teach agents what matters by turning domain knowledge into queryable context they can use to understand your repository and find what matters faster.

Less blind search. Less repeated discovery. Reusable expert intent.

View App on GitHub View CLI on GitHub

GitSense Chat builds and packages reusable repository intelligence. The gsc CLI makes it available in terminals and agent sessions.

No More Blind Searches

# Install the GitSense CLI
curl https://gitsense.com/gsc.sh | bash

# Try a smart repository
git clone https://github.com/gitsense/smart-ripgrep
cd smart-ripgrep

# Create the queryable database
gsc manifest import code-intent

# Search with repository context
gsc rg cache --db code-intent --fields purpose

Teach Agents Once

Teaching an agent with GitSense Chat starts as a natural conversation. Package the result into a Brain, then let your agent use it during real work.

Step 0: Install the CLI

Install gsc before importing Brains or initializing your agent.

# Install the GitSense CLI
curl https://gitsense.com/gsc.sh | bash

Step 1: Setup Smart Repository

Clone the example repository, then import the Rust test Brain.

# Clone the smart repository
git clone https://github.com/gitsense/smart-ripgrep

# Enter the directory
cd smart-ripgrep

# Import the Rust test Brain
gsc manifest import rust-test-coverage-intent

Step 2: Start Your Agent

Initialize your agent so it knows how to use the repository Brain.

# Start your agent (Claude Code, Codex, etc.)
claude

# In your agent session, have it run:
gsc experts init

Step 3: Ask a Real Question

Let the agent use the Brain before opening files, then ask it to explain the fallback.

First, ask for guidance from the Brain.

I want to add a regression test for binary file handling.

Use the rust test coverage brain to find the files I should look at.
Do not open the files yet. Just use the brain to guide me for now.

Then ask what the agent would need to do without it.

If the rust test coverage brain did not exist, how would you have helped me?

Teach Agents What Matters

GitSense Chat handles the hard parts: creating Analyzers, running analysis, and packaging portable intelligence your agents can query.

Create

Describe the pattern your agents should understand. GitSense turns the chat into a reusable Analyzer.

Analyze

Select files, apply filters, and run the Analyzer in batches. GitSense tracks progress and results.

Package

Combine analysis into a queryable manifest your CLI and agents can use.

Portable Intelligence

This is how GitSense makes repositories smart: intelligence is packaged as a plain JSON manifest that can live with the repository.

GitSense does not change how you store code. The .gitsense directory stays out of the way.

You own the manifest. Commit it, publish it, import it by URL, or keep it private. GitSense does not lock you in.

repo/
  .gitignore
  README.md
  src/
  .gitsense/
    .gitignore
    manifests/
      code-intent.json
      ownership.json
      review-rules.json

Get Started

Install the CLI, start GitSense Chat, and ask your agent for help from inside a repository.

Use the app to build and package repository intelligence. Use the CLI to query it from your terminal and share it with agents.

View App on GitHub View CLI on GitHub

Quick Start

# Install the CLI
curl https://gitsense.com/gsc.sh | bash

# Install and start GitSense Chat App
gsc app native install
gsc app native start

# In your agent session
gsc docs help