Skip to content

Outfit your team.

Roll out a team of agents for you or your team. AI Outfitter is open source—use our GitHub App to jump-start adoption, or check out the setup docs!

Agentic workflows

Set up a personal or organization .agents repository with our GitHub App or configure it yourself. Adopt a workflow to give your agents shared context, skills, and instructions, then put them to work across your team.

01 / SETUP

Set up your team

Use the GitHub App to create or manage your .agents repository, or follow the manual setup guide.

02 / ADOPT WORKFLOWS

Choose how your agents work

Select a ready-made workflow and inspect the context, skills, and instructions it provides.

03 / START SHIPPING

Put your agents to work

Use the shared workflow across your team’s supported agent tools and start applying it to real work.

Rendering workflow…

Workflow step

Start your first software factory → · Explore the workflow →

Rollout

Keep agent context, skills, tools, and permissions in a reviewable .agents/ tree. Give one person an agent team—or roll it out across your organization.

01 / OUTFIT

Specialists for every function

Compose agents for research, marketing, business development, engineering, and operations.

02 / ROLL OUT

From one person to every team

Publish shared skills and context once. Let each team adapt them to its work.

03 / DELEGATE

Work where your team works

Delegate from a coding harness, GitHub, CI, or a resident agent on your infrastructure.

Publish once. Delegate from every harness.

Section titled “Publish once. Delegate from every harness.”

An admin publishes the organization’s shared agent context once. A non-admin employee syncs it, then Pi and Claude Code receive the same delegation prompt: turn bounded work into a durable issue and let issue triage assign it to a resident agent. Managed resident deployment and hosted inference are planned capabilities; the dashboard labels them as planned rather than available.

Rendering workflow…

Workflow step

Follow the handoff through issue triage and the software factory →

The adoption ramp

Start with one delegated task. Share the configuration that worked. Move it to an automated surface. Add governance and durable evidence when the system earns more authority. Each rung keeps the useful parts of the last.

AssistedDelegatedAutomatedGovernedSelf-improving

See how to move the control point outward →

Open by construction

Bring your harness. Keep your configuration.

Section titled “Bring your harness. Keep your configuration.”

Outfitter does not invent a private configuration format. The .agents/ directory remains useful on its own and can target Pi, Claude Code, Codex, and the execution surfaces your organization already operates.

Terminal window
npx @ai-outfitter/outfitter

Read the getting started guide →