Grits
A Git-native, local-first issue tracker designed for AI Agents and humans.
Status: v2.9.0 — "Agent Automation"
Quick Start
Prerequisites
- Rust (latest stable)
- Git
Install CLI
# Via Cargo (recommended)
# Or download binary from Releases
# https://github.com/babybirdprd/grits/releases
Initialize a Project
Core Workflow
1. Check Pulse
Hydrate your session (or your agent's context) with the current project status.
2. Create Issue
3. Work on Issue
Sets the "sticky focus" and updates status to in-progress.
4. Context Awareness (Star Neighborhood)
Find files topologically connected to your target.
5. Assemble Context
Bundle relevant code into a Markdown format for LLM analysis.
CLI Reference
| Command | Purpose |
|---|---|
gr list |
List issues (filter by status, assignee, etc.) |
gr show <id> |
Show issue details |
gr update |
Update issue fields |
gr close <id> |
Close an issue |
gr pulse |
Session hydration (Focus + Blockers + Recent Commits) |
gr workon <id> |
Set focus and status |
gr star <file> |
Get connected files (imports/calls) |
gr context assemble |
Generate mini-codebase from focus or symbols |
gr export |
Export to JSONL (for git sync) |
gr import |
Import from JSONL |
gr stats |
Simple issue statistics |
Agent Skills
Grits exposes its functionality as a standardized "Agent Skill" for AI agents. This allows agents to natively discover and utilize Grits for task management and context retrieval.
- Skill Definition: .agent/skills/grits/SKILL.md
- Standard: Follows the Agent Skills open specification.
To use this skill with a compatible agent (e.g., in a browser environment or agent runner), point the agent to the .agent/skills/grits directory.
Documentation
Architecture
Twin Engine:
- SQLite (
.grits/grits.db) — Fast local queries. - JSONL (
.grits/issues.jsonl) — Git-versioned source of truth.
Graph-Lite:
- Uses a lightweight AST parser to build a dependency graph.
- Provides "Star Neighborhoods" (connected files) to agents without heavy analysis.
License
MIT