Skip to content

Factories > Factory agents

Configuring factory agents

Configure a factory agent's instructions, tools, model, and harness, including authentication for Claude Code, Codex, and Pi.

Open in ChatGPT ↗
Ask ChatGPT about this page
Open in Claude ↗
Ask Claude about this page
Copied!

Change how an existing factory agent works by editing its instructions, tools, and runtime in the Warp Factories web app or its definition files.

If your factory definition is hosted by Warp, edit agent settings in the factory web app. If the definition is hosted by GitHub, settings are read-only in the app; edit the definition files in your repository instead.

  1. In the web app sidebar, click your factory to expand it, then click Agents.

    The Agents page in the factory dashboard, listing a foreman agent and its four default agents: triage, spec, implement, and review.

    The Agents page lists a factory’s foreman and default agents.

  2. Click an agent to open its Settings tab.

    An Implement agent's Settings tab showing its description, MCPs, secrets, and Harness, Model, Host, and Runner dropdowns.

    The Settings tab contains the agent’s description and resources.

  3. Edit the agent description, harness, model, runner, host, MCP servers, or secrets.

  4. To change the agent’s instructions, click Prompt and edit the prompt.

  5. In the top-right corner, click Save.

Choose each agent’s model and harness in its settings. Supported harnesses include the Warp Agent, Claude Code, Codex, and Pi. A foreman using Claude Code or Codex can still dispatch other factory agents, whose runs are tracked as its children.

Third-party harnesses are included on the Scale and Enterprise plans for Warp Factories. Existing paid-plan access depends on your team’s plan and workspace settings. See plan requirements.

Default model IDs change over time, so choose based on what each agent has to do well:

AgentWhat to optimize for
ForemanOrchestration, instruction following, and long-running conversations
TriageResearch, evidence gathering, and working with connected tools
SpecSynthesizing requirements, technical reasoning, and precise writing
ImplementCoding strength, with a harness that fits your repositories and toolchain
ReviewA different model or harness from the implement agent, so the two don’t share blind spots

The model picker also includes custom routers from factories you can access. See Agent model choice, Custom routers, and Agent harnesses for available options.

For team-managed inference in definition files, see custom endpoint and BYOK model values, including factory defaults and per-agent overrides.

Use factory ZDR guidance to check coverage for your model and provider account before changing settings. For Warp-managed inference, choose models outside the provider-retention exception list.

Review factory defaults, agent and automation overrides, and separate feature models such as Computer Use and Scorers. The factory definition reference lists these settings. Team-wide model restrictions and shared credentials belong in the Admin Panel’s Models settings.

Choose the path matching where your factory definition is hosted.

With editing access:

  1. In the Warp Factories web app, expand your factory and click Factory definition. Review model defaults and overrides in the existing definition files.
  2. In Agents, open each agent’s Settings, including the foreman’s, and review the “Harness” and “Model” selections.
  3. Save any changes, resolve validation errors, and reopen the settings to confirm the intended values before checking a new run.

You need permission to propose and merge definition changes. The factory web app shows these files read-only.

  1. In the Warp Factories web app, expand your factory, open Settings, and follow the definition repository link.
  2. In GitHub, propose model changes in the existing definition files. Check factory defaults, agents including the foreman, and any automation or feature overrides.
  3. Confirm the factory-definition PR checks pass, then merge into the connected production branch. Reopen the changed settings in the factory web app to confirm the values before checking a new run.

Use a test factory with non-sensitive repositories and inputs:

  1. In the factory web app, click New Chat and choose the test factory below the prompt input.
  2. Send: “Reply with a greeting. Don’t read repository files or call tools.”
  3. In Runs, open the new run and compare its model and harness with your selections. Check child runs and any features you changed separately.

If settings don’t match, review overrides or contact support before sending sensitive work. A greeting checks the selected model, not coverage for every AI request.

Change an agent’s harness to Claude Code or Codex to run it with that provider’s own coding tool. For Pi setup, see Configuring the Pi harness.

  1. In the web app sidebar, click your factory to expand it, then click Agents.

  2. Click an agent to open its Settings tab.

  3. In the “Harness” dropdown, choose Claude Code or Codex. The “Auth” field appears for these harnesses.

  4. In the “Auth” field, choose a compatible, team-owned secret. To create one, click New auth secret for Claude Code or New OpenAI API Key for Codex.

    An Implement agent's Settings tab with Claude Code selected in the Harness dropdown and the Auth dropdown open, showing a saved credential and New auth secret.

    The Auth dropdown appears when a third-party harness is selected.

  5. In the “Model” dropdown, choose a model from that harness’s catalog.

  6. In the top-right corner, click Save.

If your factory definition is hosted by Warp, edit the harness, auth, and model in the factory web app. If the definition is hosted by GitHub, edit agentDefaults.harness or a per-agent harness override in your repository. For an example, see 03-multi-harness.