AI-Pack Documentation

AI-Pack is a Claude Code plugin that turns a single Claude Code session into a coordinated engineering team.

What is AI-Pack?

AI-Pack ships three things:

• Seven specialized subagents — architect, engineer, inspector, pr-shepherd, product-manager, reviewer, and spelunker. Each is a self-contained role definition with its own tool discipline, quality gates, and reporting format.
• Four workflow skills — orchestrate (decompose and delegate multi-step work), prd (product-discovery interview producing a PRD), pre-push (review-and-fix loop on local commits), and shepherd-pr (drive a GitHub PR to a green, approved state).
• A knowledge-graph MCP server (kg) — persistent, per-project memory that lets agents accumulate and recall findings across sessions.

Who is it for?

Anyone using Claude Code on real software projects who wants more than a single generalist assistant: parallel specialists for implementation and review, a repeatable path from "fix this bug" to a merged PR, and project knowledge that survives session boundaries.

How it works

Claude Code provides the execution loop — the tools, the permission system, and the native Agent tool for spawning subagents. AI-Pack provides what runs on top of it:

• Roles. Each subagent definition constrains a spawned agent to one job (design, implement, investigate, review, shepherd) with clear quality gates and a structured completion report.
• Coordination. The skills package proven multi-agent patterns: the orchestrate skill decomposes work and delegates it to the right roles, with parallel spawns where tasks are independent; pre-push and shepherd-pr run bounded fix-and-verify loops.
• Memory. Subagents share no context with each other or with your main session — every brief must be self-contained. The kg knowledge graph fills the gap: agents checkpoint findings into it and pull prior context out of it, so knowledge persists across agents and across sessions.

Documentation structure

• Getting Started — install the plugin and verify it works
• Agents — the seven subagents and when to use each
• Skills — the four workflow skills and what triggers them
• Knowledge Graph — persistent project memory via the kg MCP server
• Task Packets — an optional convention for briefs that outlive a session
• Workflows — general engineering process guides: bugfix, feature, refactor, research, standard
• Clean Code — coding standards and best practices agents are held to

Support

• GitHub Issues
• GitHub Discussions

History

AI-Pack 1.x/2.0 was an API-driven agent server: a Go server running coding agents against the Claude API, with an agent CLI, an agent-mcp MCP server, and a React GUI. That architecture was deprecated on 2026-08-22 in favor of the plugin model — Claude Code natively provides the execution loop the server used to implement. The server-era code is preserved at tag v2.0-server-final.