Several agents draft. Then they review each other.

When to use this pattern#

Use this when you want several approaches to the same writing task before choosing a direction. It works for tutorials, proposals, and other documents where structure and explanation matter.

Why this division of work helps#

• Compare different approaches. You can see different ways to organize the material and explain it to your audience.
• Have writers check each other’s work. Peer reviews can reveal unclear steps, missing information, and claims that need evidence.
• Revise, then review again. Each writer revises their own draft using the feedback. The next round checks whether those changes address the issues.
• Try different models. Different models may suggest other ways to explain the subject. Separate sessions with the same model can also produce drafts worth comparing.

Writing and reviewing several drafts adds time and model usage. Use this when comparing approaches matters; one writer is usually enough for a straightforward document.

1. Start the coordinator#

• Open the project where the writers will work.
• Keep the reference material where the writers can read it.

The table shows one example setup. You can use other installed CLIs and models available to your accounts.

In your project, open Facade, click New Agent, and choose Claude Code for the coordinator. Set the model with /model opus and effort with /effort medium.

You only need to start the coordinator manually. It launches the other agents for you through Facade. If this is your first time using Facade, check out the quickstart.

2. Send your task#

Send this prompt to the coordinator. Replace the first line with your topic, audience, references, and the folder where the drafts should be saved:

Write about [topic] for [audience], using [key facts or references]. Save the drafts in [output folder].

Use Facade to start three writers in this project:
- Claude Code with opus and high effort
- Codex with gpt-6-sol and high effort
- Antigravity with gemini-3.8-flash-high and high effort

Give them the same brief and a separate draft file each in the output folder.
Have them write independently, then review each other’s drafts.
Repeat the reviews and revisions until the writers agree on the facts and the drafts meet the brief.

3. How the team works#

1. Share the brief
2. Write independently
3. Review each other’s drafts
4. Revise and review again

1. The coordinator gives each writer the same brief and reference material, with a separate draft file in the output folder.
2. Each writer completes a first draft before reading the others’ work.
3. The coordinator arranges the cross-review. Writers check claims against the reference material and point out unclear steps, missing information, or inaccuracies.
4. Each writer revises their own draft using the feedback. They repeat the reviews until they agree on the facts and the drafts meet the brief.

The output folder contains a reviewed draft from each writer. You can ask the coordinator to compare the versions. If the writers cannot resolve a disagreement, the coordinator brings their different views back to you.

Use it in your project#

• Use the same process for tutorials, proposals, product specs, or architecture documents.
• Tell the writers what matters most in the review, such as technical accuracy, useful examples, or clear explanations.
• Adjust the number of writers to your task. Two writers can be enough for a smaller comparison; you can also use the same model in separate sessions.

You can change the CLIs, models, and effort in the prompt. For saved launch settings and other providers, see Agent settings, Claude Code with another provider, and Codex with another provider.