Oh My Subagents 🤖 Agent Free

What it does

Oh My Subagents (oms) is a local orchestration layer for running Claude Code and Codex as coordinated teams of subagents. Instead of ad-hoc delegation where a lost connection or closed browser tab means starting over, it commits assignments and "waves" of work to durable state so a run can be resumed exactly where it left off, without an agent fabricating a false "completed" status along the way.

What is inside

• A controller that owns task state and only accepts explicit completed/blocked results from task leads, rather than trusting free-form agent claims
• Reusable Workflow definitions: responsibility trees you can publish once and reuse across missions, instead of hand-wiring delegation every run
• A visual Workflow Studio for designing and publishing these responsibility trees on a canvas
• An integrated console showing team structure, live activity logs, pending human requests, and workspace file references
• Eight provider-neutral starter workflows covering common scenarios like feature delivery, incident investigation and security audits
• SQLite storage by default (optional PostgreSQL), so there's no external database dependency to run it

Works with

Claude Code and Codex are the primary supported task providers; workflows are written to be provider-neutral where possible.

How to install or connect

pipx install oh-my-subagents
oms init
oms service install

Requires Python 3.12+ on Linux, macOS 13+, or Windows 11 x64.

Maintenance and safety

Actively developed (144+ commits on main) with documentation and example workflows; MIT-licensed for the core, while the bundled visual Console reuses n8n-derived UI code under the Sustainable Use License (free for internal/non-commercial use, restricted redistribution) — check that license before redistributing the Console itself. Subagents do not inherit parent capabilities, and assignment ownership is immutable, which limits accidental privilege creep.

Who should use it

Teams or solo developers running Claude Code or Codex on multi-step tasks who need delegation that survives interruptions and want an auditable record of what each subagent actually did, rather than a one-shot orchestration script.