Swarma provides a GROWS experiment loop for agent teams: generate hypotheses, run experiments, observe signals, weigh verdicts, and stack validated patterns into a playbook. It includes CLI commands, team templates, and tooling for logging metrics and running continuous cycles.
Use swarma when you want to run systematic A/B tests or growth experiments at scale with agent teams (acquisition, activation, retention, revenue). Best when you need repeatable experiment cycles and automated verdicting.
swarma init, swarma cycle, swarma run) and server modes for REST/MCP integrationTerminal-capable agents, Claude Code, Hermes, and environments that can run Python CLI tools.
Swarma is a growth experiment framework using agent teams that run A/B tests and build validated playbooks via a GROWS loop (Generate, Run, Observe, Weigh, Stack). No bundled scripts were present for runtime testing. SKILL.md is comprehensive with clear onboarding, 18 pre-built squad templates, MCP tool reference, and troubleshooting. Requires OpenRouter API key for LLM calls.
No security concerns. Well-structured SKILL.md with no scripts bundled. All API key handling is via environment variables, no hardcoded secrets. The skill directs users to install a pip package which is standard practice.