Skip to main content

Introduction to ACE

ACE is a Playbooks as a Service solution that provides self-improving AI instructions for your agents. Built on the Agentic Context Engineer (ACE) architecture, the platform helps your AI agents get better at their tasks over time.

What is ACE?​

ACE stands for Agentic Context Engineer. It's a three-agent architecture that continuously improves playbooks based on real-world outcomes:

  1. Generator - Produces outputs based on playbook instructions
  2. Reflector - Analyzes outcomes to identify improvement opportunities
  3. Curator - Synthesizes feedback into improved playbook versions

What are Playbooks?​

Playbooks are structured instructions that guide AI agents on how to perform specific tasks. Unlike static prompts, ACE playbooks:

  • Evolve automatically based on recorded outcomes
  • Version controlled so you can track changes over time
  • Accessible via MCP for seamless integration with AI agents that support MCP

Key Features​

Self-Improving Instructions​

Record outcomes from your AI tasks, and ACE automatically improves the underlying playbooks. The more you use them, the better they get.

MCP Integration​

Access your playbooks directly from Claude Desktop, Claude Code, or any MCP-compatible agent. No API integration required.

Version History​

Every evolution creates a new version. Compare changes, understand improvements, and roll back if needed.

Usage Analytics​

Track how your playbooks are being used and monitor evolution progress through the dashboard.

Quick Example​

Use the ace record_outcome tool with:
- playbook_id: "abc123"
- task_description: "Summarized quarterly earnings report"
- outcome: "success"
- notes: "Summary was accurate and well-structured"

After enough outcomes are recorded, ACE automatically evolves the playbook to incorporate lessons learned.

Getting Started​

Ready to try ACE? Here's the fastest path:

  1. Create an account - Sign up and verify your email
  2. Quick Start - Set up your first playbook in 5 minutes
  3. Core Concepts - Understand playbooks, outcomes, and evolution

Use Cases​

ACE is ideal for:

  • Code Review Agents - Improve review quality based on feedback
  • Documentation Writers - Learn from corrections and preferences
  • Data Analysis - Refine analysis approaches from outcomes
  • Customer Support - Enhance response quality over time
  • Content Generation - Adapt to style and quality feedback

Architecture Overview​

┌─────────────────────────────────────────────────────────────┐
│                           ACE                                │
├─────────────────────────────────────────────────────────────┤
│                                                              │
│  ┌──────────┐    ┌──────────┐    ┌──────────┐              │
│  │Generator │───▶│Reflector │───▶│ Curator  │              │
│  └──────────┘    └──────────┘    └──────────┘              │
│       │                               │                     │
│       ▼                               ▼                     │
│  ┌──────────┐                   ┌──────────┐               │
│  │ Playbook │◀──────────────────│ Evolved  │               │
│  │  v1.0    │                   │  v2.0    │               │
│  └──────────┘                   └──────────┘               │
│                                                             │
└─────────────────────────────────────────────────────────────┘

Next Steps​