APM Overview
APM (Actions Per Minute) is a mission control system for orchestrating AI-augmented workflows. It provides a persistent, game-inspired interface for teams to manage tasks, coordinate AI agents, and handle rapid decision-making across distributed workflows.
The Vision
In a world where teams are managing fleets of AI agents and processing hundreds of outputs simultaneously, traditional project management tools fall short. APM applies proven UX patterns from competitive gaming, where split-second decisions and high-cognitive load are the norm, to create an interface that keeps humans effectively in the loop.
Core Architecture
APM as MCP Client
APM as MCP Server
APM is built entirely around the Model Context Protocol (MCP), functioning as both:
- MCP Server - Exposing APM's functionality to external AI agents and tools
- MCP Client - Connecting to and orchestrating external MCP servers
This dual architecture means all external interactions with APM happen through MCP - there's no separate API to learn.
KeyConcepts
Projects
The top-level container for organizing work. Projects group related tasks and store project related settings for agents to act on.
Tasks
Atomic units of work that can be assigned to humans or AI agents. Tasks flow through a simple lifecycle (Unstarted -> In Progress -> Complete) and can trigger MCP events for automation
Actions
Decision points requiring human judgment. When AI agents need approval, review, or validation, they create Actions that appear in APM's real-time interface for rapid processing
MCP Integration
The nercous system connecting everything. MCP servers can: create and update tasks, submit actions for approval, query project state, react to events like task completion
Why Game-Inspired?
We're not gamifying work, we're borrowing UI patterns that already solve for:
- High-frequency decision making - process 100+ microdecisions efficiently
- Real-time state awareness - Know what's happening across all agents/tools
- Persistent overlay interface - Stay in context while switching between tasks
- Visual priority systems - Instantly identify what needs attention
Getting started
Next Steps
- Understand Core Concepts - Deep dive into Projects, Tasks, and Actions
- MCP Integration Guide - Learn how to connect MCP servers
- Build Your First Integration - Create a custom MCP Server
Example Use Cases
Software Development
- AI writes code -> Creates action for review -> Developer approves -> Task marked complete
- Multiple AI agents work on features -> APM coordinates dependencies -> Team reviews in parallel
Content Operations
- AI generates articles -> Editorial actions created -> Editors review/approve -> Auto-publish on approval
- Research agents gather data -> APM organizes findings -> Writers enhance with human insight
Customer Support
- AI processes tickets -> Complex issues become tasks -> Support team handles escalations
- Sentiment analysis creates high-priority actions -> immediate human intervention when needed