Oh My OpenAgent — The Best Agent Harness | Oh My OpenAgent

Star on GitHub

The Best Agent Harness<br>58.6k stars. 2.5M+ downloads. Battle-tested and production-proven.

58.6k GitHub Stars<br>11 Specialized Agents<br>2.5M+ Total Downloads<br>519.1k Monthly Downloads<br>54+ Lifecycle Hooks

$bunx oh-my-openagent install

Get StartedGitHubView on GitHub

The Magic Wordulw<br>Type Three Letters. Walk Away.<br>Ultra Work mode triggers maximum precision: auto-planning, deep research, parallel agents, and self-correction loops. The system doesn't stop until it's done. You don't babysit.

Auto-PlanningDeep ResearchSelf-CorrectionParallel Agents<br>Zero intervention. Full autonomy. Just results.

sisyphus — -zsh — 80x24

➜~_<br>Scanning codebase...<br>Launching 3 explore agents...<br>Building execution plan...<br>Deploying 4 parallel agents...<br>Running verification suite...

✓Done. 847 lines changed, 0 issues.<br>➜~_

Primary OrchestratorClaude Opus 4.7 Max<br>Sisyphus<br>The CTO Who Never Sleeps<br>Named after the mythological figure rolling his boulder uphill forever. Sisyphus parses implicit requirements, adapts to your codebase's maturity, and delegates to specialists. Interrupted? The boulder system resumes exactly where you left off.<br>PHASE 1<br>Intent Gate

Parses what you meant, not just what you typed

PHASE 2<br>Codebase Assessment

Maps your architecture before touching a line

PHASE 3<br>Smart Delegation

Routes to the right specialist agent

PHASE 4<br>Independent Verification

Trusts nothing. Verifies everything.

Session Continuity<br>Active work tracked in boulder.json. Power outage? System crash? Doesn't matter. Pick up exactly where you stopped.

Orchestration WorkflowThink, Then Act<br>Planning and Execution Are Different Jobs

Claude Opus 4.7 Max<br>Prometheus<br>Strategic Planner

The architect. Interviews you, explores your codebase, and creates detailed battle plans. Never writes code. Uses Metis and Momus as quality gates.<br>Intelligent interview mode<br>Multi-agent codebase exploration<br>Gap analysis with Metis<br>Ruthless review with Momus

Claude Sonnet 4.6<br>Atlas<br>Master Executor

The builder. Reads verified plans, delegates to specialized agents via category+skills system. Tracks learnings across tasks. Verifies independently — never trusts subagent claims.<br>Intent-based task routing<br>Wisdom accumulation across tasks<br>Boulder session continuity<br>Independent result verification

Describe your work to Prometheus

Metis analyzes for gaps

Momus validates ruthlessly

Atlas delegates to specialists

Independent verification on everything

The planner doesn't code. The executor follows verified plans. Intelligence lives in the system, not individual agents.

Deep WorkerGPT 5.5 Medium<br>Hephaestus<br>Give Him a Goal, Not a Recipe<br>Named after the Greek god of the forge. Methodical, thorough, obsessive. Fires 2-5 parallel explore agents before writing any code. Built for deep architectural reasoning, complex debugging, and cross-domain synthesis.<br>01<br>EXPLORE — Map the terrain

02<br>PLAN — Chart the course

03<br>DECIDE — Commit to the path

04<br>EXECUTE — Build with precision

05<br>VERIFY — Prove it works

For when good enough isn't.

NEW · v4.0opt-in<br>Team Mode<br>Multiple Agents. Side-by-Side. In Parallel.<br>A lead agent orchestrates a team of category-specialized members, all running in parallel and communicating through dedicated tools. Watch every member work simultaneously in tmux. Opt-in — flip team_mode.enabled and restart.

Lead Orchestration

One coordinator. Multiple specialists. Real-time messaging via team_send_message and team_task_create.

True Parallelism

Up to 8 members work simultaneously. No waiting, no queues — every angle covered at once.

Tmux Visualization

Focus and grid windows let you zoom into one teammate without losing the bigger picture.

team_* Tool Family

12 dedicated tools (team_create, team_status, team_send_message, ...) gated behind team_mode.enabled.

Powered by Team Mode

hyperplan

5 hostile agents tear apart your plan from orthogonal angles before a single line of code is written.

security-research

3 vulnerability hunters + 2 PoC engineers audit your codebase. Severity calibrated by actual exploitability.

team_mode.enabled = true → restart opencodeOff by default. Devastating when on.

Specialized Agents<br>Five built-in agents for common workflows. Plus a dynamic system that assembles custom agents on demand.

GPT 5.5 High<br>Oracle<br>Architecture Consultant

Complex debugging and architectural decisions. When the path forward isn't obvious.

GPT 5.4 Mini Fast<br>Librarian<br>Docs & Code Search

Finds real GitHub examples and official docs. Evidence-based, with permalinks.

GPT 5.4 Mini Fast<br>Explore<br>Codebase Grep

Blazing fast codebase search. Cheap, parallel, always background.

Claude Sonnet 4.6<br>Metis<br>Plan Consultant

Catches ambiguities in plans before they become production bugs.

GPT 5.5 xHigh<br>Momus<br>Plan Reviewer

Ruthless validation. Only approves when the plan is bulletproof.

Assembled for the task at hand<br>Dynamic Agent<br>Matches your request to the right model, loads...