# AI use cases for Anneal (mosonlab/anneal)

Repo: [mosonlab/anneal](https://github.com/mosonlab/anneal) | Use cases: 4 | Page: https://dedupedai.com/use-cases/mosonlab-anneal/

Anneal is a local system that runs chains of coding agents to plan, review, implement, verify and merge tasks queued on a board, using the Codex and Claude subscriptions you already have.

This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL.

- Deliver queued specs as merged pull requests overnight: https://dedupedai.com/use-cases/mosonlab-anneal/deliver-queued-specs-as-merged-prs/
- Trace which commits an Anneal chain produced: https://dedupedai.com/use-cases/mosonlab-anneal/trace-chain-history-from-git-log/
- Reshape the twelve-step chain into your own agent workflow: https://dedupedai.com/use-cases/mosonlab-anneal/reshape-chain-into-custom-agent-workflow/
- Run chains for several projects at the same time: https://dedupedai.com/use-cases/mosonlab-anneal/run-several-projects-in-parallel/

## How to deliver queued specs as merged pull requests overnight with Anneal

Repo: [mosonlab/anneal](https://github.com/mosonlab/anneal) | Difficulty: Advanced | Time: 2 hours | Basis: From the README

Write task specs, queue them on the board, and collect merged pull requests when the chains finish.

**Who it is for:** A solo developer or small team lead with Codex or Claude subscriptions who has a backlog of well-written specs and wants them implemented without babysitting each run.

**The problem:** Turning written specs into reviewed, tested code takes many manual handoffs between planning, coding, review and merging.

**The outcome:** Each queued task moves through plan, review, implementation, two code reviews, fixes, regression checks and merge, and the board shows merged pull requests and any items waiting on a human answer.

### Steps

1. Clone the repository and check out the release: `git clone https://github.com/mosonlab/anneal.git`, then `git checkout v0.9.0`.
2. Install and build: `npm ci`, `npm run setup:local`, `npm run build`.
3. Start the database and migrate: `docker compose up -d --wait --wait-timeout 60 postgres`, then `npm run db:migrate:release -- --fresh`.
4. In three terminals, run `npm run dev:api`, `npm run dev:runner` and `npm run dev:web`, in that order, then open `http://127.0.0.1:5173`.
5. Add your project following the pull-request workflow guide, write specs, queue them on the board, and check the Inbox for agent questions.

### Prompt to paste into your agent

```text
Using the mosonlab/anneal repository, set up the local release at v0.9.0 and queue the spec in <path-to-spec-file> on the board for <project-name>. Report the board status and the pull request URL when the chain finishes. Verify that the chain's merge step completed and that the pull request is merged into the target branch.
```

### Caveat

The README says the only upgrade path is a fresh install and interfaces may change between previews, so this is a developer preview for v0.9.0 and not a stable production tool.

Risk, credentials: Anneal runs the CLIs non-interactively as your own user with your access, so chains can change code and merge to main without a permission prompt; test on a throwaway repository first.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/mosonlab-anneal/deliver-queued-specs-as-merged-prs/

## How to trace which commits an Anneal chain produced with Anneal

Repo: [mosonlab/anneal](https://github.com/mosonlab/anneal) | Difficulty: Intermediate | Time: 1 hour | Basis: From the README

Find chain-delivered commits in git history and follow each step through its recorded artifacts.

**Who it is for:** A reviewer or maintainer who needs to know which commits in a repository were written by Anneal chains, and why a change was merged.

**The problem:** Chain-generated commits look like normal commits, so it is hard to tell which changes passed through planning, review and regression checks.

**The outcome:** A list of commits carrying the Anneal trailers, with run and step identifiers, that can be matched to recorded plan, review and regression artifacts.

### Steps

1. Open the repository and confirm the Anneal trailers are present: `git log --grep="X-Anneal-Run"`(inferred). (inferred, check the README)
2. Note the `X-Anneal-Run` and `X-Anneal-Step` values on each relevant commit.
3. Look up the matching run in the local Anneal board to read its plan, review findings, fix dispositions and regression results.
4. Compare the merged pull request with the recorded artifacts to confirm which review findings were adopted.

### Prompt to paste into your agent

```text
In the <repo-path> repository, list every commit with an X-Anneal-Run trailer and group them by run identifier. Return a table of run, step and commit hash, and verify that each run identifier appears in the Anneal board records.
```

### Caveat

The README states the trailers and recorded artifacts exist, but it does not describe how to browse the artifacts, so the lookup step depends on the local board interface.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/mosonlab-anneal/trace-chain-history-from-git-log/

## How to reshape the twelve-step chain into your own agent workflow with Anneal

Repo: [mosonlab/anneal](https://github.com/mosonlab/anneal) | Difficulty: Advanced | Time: 3 hours | Basis: From the README

Edit agent roles, prompts, models and gates in the template data to fit your own team of agents.

**Who it is for:** A team lead or engineer who wants the chain to use different review rules, roles or models than the built-in Full Assurance template.

**The problem:** A fixed delivery chain may not match how a team reviews code, splits work or chooses models.

**The outcome:** A customized template with your own roles, prompts and gates, used by new chains on the board.

### Steps

1. Read the template files in `agents/templates/compound-engineer-workflow/` to see how each role binds a prompt, model, reasoning effort and runner.
2. Read the task routing contract in `docs/governance/task-routing-v1.md` to learn what each board column means.
3. Copy the template, then change roles, prompts, models or gates for the steps you want to alter.
4. Queue a small test task on a throwaway repository and confirm each step runs the role you set.

### Prompt to paste into your agent

```text
In the mosonlab/anneal repository, copy agents/templates/compound-engineer-workflow to a new template named <template-name>, change the step that uses the role <role-slug> to use <model> at <effort>, and report the changed fields. Verify that the template still lists twelve steps in order.
```

### Caveat

The README says roles and gates are editable but does not document the exact file format or how a new template is registered, so those details must be read from the repository.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/mosonlab-anneal/reshape-chain-into-custom-agent-workflow/

## How to run chains for several projects at the same time with Anneal

Repo: [mosonlab/anneal](https://github.com/mosonlab/anneal) | Difficulty: Advanced | Time: 2 hours | Basis: From the README

Queue tasks across multiple repositories and let independent chains run in parallel on one machine.

**Who it is for:** A developer maintaining several repositories who wants backlog items moving in each project at once.

**The problem:** Working through several repositories one chain at a time leaves most of the machine and subscription capacity idle.

**The outcome:** Chains for different projects and tasks progress in parallel, limited only by machine resources and subscription rate limits.

### Steps

1. Add each repository as a project following the pull-request workflow guide.
2. Queue one or more specs per project on the board.
3. Register additional runners if you want more throughput, keeping within your machine and subscription limits.
4. Watch the board and answer Inbox questions as they appear, since each chain pauses until a human decision is given.

### Prompt to paste into your agent

```text
Using the mosonlab/anneal repository, queue one spec for each of the projects <project-a> and <project-b> on the board, then report which chains are running at the same time. Verify that both chains advance without waiting on each other.
```

### Caveat

The README gives no concurrency numbers, so the practical limit depends on your own machine and your provider plan limits.

Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/

URL: https://dedupedai.com/use-cases/mosonlab-anneal/run-several-projects-in-parallel/