Manage complex projects with a team of AI agents, smoothly and efficiently.

APM is an open-source framework for managing ambitious software projects with AI assistants. Instead of working in a single, increasingly chaotic chat, APM structures your work into a coordinated system where different AI agents handle planning, coordination, and execution as a team.

What distinguishes this from subagent-based approaches: the agents doing implementation work are not restarted fresh on each task. They accumulate working context across assignments - building familiarity with their domain as the project progresses. When context fills, a structured Handoff transfers that working knowledge to a fresh instance rather than discarding it.

As conversations grow, AI context degrades. The assistant loses track of requirements, produces bad code, and hallucinates details. For substantial projects, this makes sustained progress nearly impossible.

To address this, APM coordinates three specialized agent types, each operating in its own context with only the information it needs:

• Planner - Conducts structured project discovery and decomposes requirements into three planning documents: a Spec (what to build), a Plan (how work is organized), and Rules (how work is performed).
• Manager - Coordinates execution by assigning Tasks to Workers, reviewing completed work, and maintaining project state. Operates on execution summaries rather than raw code.
• Workers - Execute Tasks within defined domains (frontend, backend, API, etc.). Each Worker receives a self-contained Task Prompt for each assignment, executes, validates, logs to memory, and reports back.

Project state lives in structured files outside any agent's context. Because of this, when an agent reaches its limits, you can Handoff its working knowledge to a fresh instance of that same agent in a new chat. This also allows completed sessions to be archived and their context carried forward to new ones.

You mediate every exchange between agents by running commands in the appropriate conversation. This keeps every step visible and auditable, letting you set the pace and review work at each stage. Each agent tells you exactly what to run, in which conversation, and what to do next.

You start the Planner with an initiation command and it asks you structured questions while reading through your codebase - you can correct, steer, add context, and sign off on its understanding. That conversation produces three planning documents that govern everything that follows.

From there it's simple: the Manager coordinates execution using the planning documents and tells you what command to run next and where, you run it in the right conversation, and work gets done. One command delivers a task to a Worker. Another brings the Worker's report back to the Manager. At each step you can proceed, iterate, adjust, or redirect.

As the project grows, Workers build up focused working knowledge of their domains. When one fills its context, you run a Handoff command and the next instance picks up without gaps. The same applies to the Manager.

APM is for people who build with AI agents and own what they ship. Delivering each message between agents is a built-in checkpoint; you see every task assignment before it reaches a Worker and every result before the Manager acts on it. Workers run in separate conversations, giving you full visibility into their work and room to interact, redirect, or course-correct at any point. The workflow is transparent by design.

APM supports Claude Code, Codex CLI, Cursor, GitHub Copilot, Antigravity, and OpenCode.

Install the CLI:

npm install -g agentic-pm

Navigate to your project directory and initialize:

apm init

Select your AI assistant when prompted. The CLI installs commands, guides, skills, and project artifact templates into your workspace.

Next, open your AI assistant and run:

/apm-1-initiate-planner

You can also provide context about what you want to build:

/apm-1-initiate-planner I want you to build Claude Opus 5. Make no mistakes.

The Planner collaborates with you through project discovery and creates the planning documents for you to review. Once approved, it guides you to open a new conversation and run /apm-2-initiate-manager to begin coordinated execution. From there, each agent directs you through the workflow step by step.

For the full walkthrough, see the Getting Started guide.

Full documentation is available at agentic-project-management.dev:

• Introduction - What APM is and how it works
• Getting Started - Installation through first task cycle
• Agent Types - Planner, Manager, and Worker roles
• Agent Orchestration - Communication, coordination, Memory, and Handoff mechanics
• Workflow Overview - Every procedure in detail

The site also covers advanced topics like how APM's prompt and context engineering works under the hood, design principles behind the multi-agent coordination, tips and tricks for model selection and cost optimization, troubleshooting, and customization.

See the CLI Guide for full details.

APM supports custom repositories for teams that want to modify the workflow. Fork the repo (for upstream sync) or use "Use this template" (for a clean start), adjust templates, build, release, and install with apm custom -r owner/repo. A customization skill is included to guide AI agents through the process.

See the Customization Guide for details.

APM Auto is an official custom adaptation of APM built for Claude Code. It replaces the user-mediated Worker model with autonomous subagent dispatch - the Manager spawns ephemeral subagents via Agent() to execute Tasks, reviews their output, and continues without requiring you to shuttle messages between chats. Best for prototyping, fast execution, and simpler projects.

apm custom -r sdi2200262/apm-auto

APM Semi is an official custom adaptation of APM for collaborative human-and-agent project execution. The User can claim any Task at any point and execute it directly while the agent on their side stays on standby - answering questions, running validation, and writing the Task Log on the User's behalf. Best for users who want to author the substantive code themselves and lean on AI for boilerplate or peripheral Tasks.

apm custom -r sdi2200262/apm-semi

The apm-assist skill turns your AI assistant into an APM-aware helper. Install it into any project and your assistant can explain how APM works, answer questions by reading the live documentation, detect your installation state and version, and guide migration from v0.5.x. It works with any supported platform.

Give this to your AI assistant to handle installation:

Install the apm-assist skill into this project. It is a general-purpose assistant skill for the Agentic Project Management (APM) framework that explains concepts, answers questions from the live documentation, detects installed versions and analyzes sessions, and guides migration from older versions to v1.

Installation instructions and platform-specific paths are in the standalone skills README:
https://github.com/sdi2200262/agentic-project-management/blob/main/skills/README.md

v1.0.0 is a ground-up redesign - the workflow, file structure, CLI, and agent roles all changed significantly. The pre-v1 codebase is preserved on the v0.5.x branch for reference.

The Troubleshooting Guide documents the recommended migration procedure. The apm-assist skill above can also walk you through it interactively.

Contributions are welcome. See CONTRIBUTING.md for guidelines. Report bugs or suggest features via GitHub Issues. Reach out on Discord: cobuter_man.

CLI and template releases version independently but share major version for compatibility. See VERSIONING.md for details.

Licensed under the Mozilla Public License 2.0 (MPL-2.0). APM is free for all uses including commercial. Improvements to core APM files must be shared back with the community. See LICENSE for full details.

Versions prior to v0.4.0 were released under the MIT license. The license was updated to MPL-2.0 starting with v0.4.0.