# ACEM

**Automatic Code Evolution Management**

ACEM brings AI-powered development directly to your GitHub workflow. No new tools to learn, no complex setup. Just create issues, add labels, and get PRs.

## How it works

ACEM runs as a background processor that continuously scans your repositories for issues with specific labels. When it finds one, it takes action:

| Policy             | What it does                                                                                         |
| ------------------ | ---------------------------------------------------------------------------------------------------- |
| `github-design`    | Reads labeled issues, browses your codebase, and creates a PR with an implementation design document |
| `github-implement` | Reads labeled issues, browses your codebase, and creates a PR with the actual code changes           |

### The workflow

1. **Create an issue** — Describe what you want to build or fix
2. **Add a label** (e.g. `acem-design`) — ACEM picks it up on the next processing cycle
3. **ACEM creates a PR** — Design PR contains a markdown implementation plan, implement PR contains actual code changes
4. **Review and iterate** — Leave review comments on the PR, ACEM picks up feedback and updates the PR
5. **Merge or request fresh start** — Design: merge when satisfied then implement. Implement: merge when ready

### Design-first workflow (recommended)

```
Issue → label: acem-design → Design PR → review → merge
     → label: acem-implement → Implement PR → review → merge
```

### Direct implementation

```
Issue → label: acem-implement → Implement PR → review → merge
```

### Iteration on designs

Leave review comments on the design PR. ACEM will regenerate the design incorporating your feedback and push an update to the same branch.

### Iteration on implementations

Leave review comments on the implementation PR. ACEM will address the feedback and push an update to the same branch.

### Starting fresh

To get a completely new implementation, close the PR without merging while keeping the label on the issue. ACEM will create a fresh PR on the next cycle.

## Powered by AI agents

ACEM uses AI agents that can read and understand your entire codebase:

* **Design agent** — Read-only access. Browses your code, understands patterns, and writes a detailed implementation plan as a markdown document.
* **Implement agent** — Full write access. Browses your code, understands patterns, and makes targeted code changes.

Both agents follow your project's conventions and coding standards.

## Pricing

Simple, transparent token-based pricing:

| Model        | Input             | Output            |
| ------------ | ----------------- | ----------------- |
| MiniMax M2.5 | $0.40 / 1M tokens | $1.50 / 1M tokens |

You only pay for what you use. Start with a design to estimate costs before implementing.

### Compute included

A VM sandbox is included with your subscription, subject to fair usage. The sandbox is used to run ACEM's code generation and testing.

Running benchmarks or keeping the sandbox busy for extended periods is not considered fair usage and may result in additional charges.

## Get started

<a href="https://github.com/RedBeardLab/acem-doc/blob/main/documentation/getting-started/quickstart.md" class="button primary">Quickstart guide</a> <a href="https://github.com/RedBeardLab/acem-doc/blob/main/documentation/concepts/workflow.md" class="button secondary">How it works</a>

***

Need help? Check our [FAQ](https://github.com/RedBeardLab/acem-doc/blob/main/documentation/faq/index.md).