Agentic Jujutsu

Agentic Jujutsu

AI agent coordination layer for Jujutsu VCS (jj) with quantum-ready architecture, QuantumDAG consensus, and AgentDB self-learning. Provides AI-powered conflict resolution, automated code review, and multi-agent VCS workflows.

Quick Command Reference

Installation

Install: npx agentic-jujutsu@latest See Installation Guide for the full ecosystem.

Core Commands

init

Initialize agentic-jujutsu in a Jujutsu repository.

npx agentic-jujutsu@latest init [--force] [--config <path>]

review

Run AI-powered code review on current changes.

npx agentic-jujutsu@latest review [--depth <level>] [--agents <n>]

resolve

AI-assisted conflict resolution.

npx agentic-jujutsu@latest resolve [--strategy <name>] [--auto]

Strategies: semantic, syntactic, llm-assisted, consensus

swarm-review

Run multi-agent swarm code review.

npx agentic-jujutsu@latest swarm-review [--agents <n>] [--perspectives <list>]

Perspectives: security, performance, style, logic, architecture

merge

AI-assisted merge operations.

npx agentic-jujutsu@latest merge [--source <branch>] [--strategy <name>]

analyze

Analyze repository history with AI.

npx agentic-jujutsu@latest analyze [--depth <n>] [--output <path>]

learn

Train AgentDB on repository patterns.

npx agentic-jujutsu@latest learn [--iterations <n>]

Common Patterns

AI Code Review

npx agentic-jujutsu@latest init
npx agentic-jujutsu@latest review --depth deep --agents 3

Conflict Resolution

npx agentic-jujutsu@latest resolve --strategy llm-assisted --auto

Programmatic API

import { AgenticJujutsu, ReviewAgent } from 'agentic-jujutsu';

const ajj = new AgenticJujutsu({ repoPath: '.', model: 'claude-sonnet-4-5-20250929' });
const review = await ajj.review({ depth: 'deep', perspectives: ['security', 'logic'] });
console.log(review.findings);

RAN DDD Context

Bounded Context: DevOps/Tools

References

• Command reference: See references/commands.md
• Full README
• npm