Open Source · v0.1 Early Preview

StrawPot

AI teams, one command away.

An open-source CLI that spins up AI agent teams—they split work, share context, and merge results, each in its own isolated workspace. Run one team or many across projects with Claude Code, Codex, OpenHands, and reusable components from StrawHub.

$ pip install strawpot
Get Started Star on GitHub
Discord Join the community GitHub View the source StrawHub Browse skills & roles

Quick Start

# Install StrawPot
$ pip install strawpot

# Start a session
$ strawpot start --role team-lead

What It Does

Isolated Sessions

Each session runs in its own sandbox—git worktree, Docker, or a custom environment. Agents work freely without touching your main workspace until you’re ready to merge results.

Multi-Agent Delegation

A lead agent breaks work into tasks, then StrawPot spawns specialists—researchers, writers, coders, reviewers—all coordinated automatically across any workflow.

Any Agent Runtime

Use Claude Code for coding, Codex for research, OpenHands for ops—or plug in your own runtime. Mix and match per role.

How It Works

StrawHub ↗

Skills, roles, agents & memory
strawpot start

Engineering

Lead Coder Tester
strawpot start

Research

Analyst Writer Reviewer
strawpot start

Marketing

Strategist Copywriter Designer
  1. Pull roles and skills from StrawHub — or define your own
  2. Run strawpot start — each session gets an isolated workspace
  3. Agents delegate sub-tasks — StrawPot resolves roles and spawns sub-agents
  4. On exit, cleans up everything automatically

Browse skills, roles, agents & more on StrawHub

StrawHub is the public registry for reusable agent components—skills, roles, agent configs, and memory providers. Discover what the community has built, publish your own, and install straight from the CLI.

$ strawpot search "code-review" $ strawpot install code-review
Explore StrawHub →

Stay in the loop

Get updates on new releases, features, and community highlights. No spam.

Powered by Buttondown