Oh My Team gives Claude Code a full development team — 12 specialized agents working in parallel across tmux panes. Plan, build, review, and debug with dedicated specialists instead of one AI doing everything. Run it locally, or deploy the Hub for always-alive multi-project sessions you control from Telegram or Slack.

Telegram Group: "Oh My Team Hub"

General topic:
  You: start ~/projects/my-app
  Hub: Started. Talk to it in its topic.

my-app topic:
  You: fix the login bug and add rate limiting
  Claude: [spawns explorer + hephaestus agents, works in parallel]
  Claude: Fixed. auth.ts updated, rate limiter added to middleware.
  
  You: run the review
  Claude: [spawns 5 review agents: goals, QA, quality, security, context]
  Claude: REVIEW PASSED. 0 blocking issues, 2 minor suggestions.

Or use it locally — no Telegram required:

omt -d
> /oh-my-team:team Build an authentication system with OAuth and RBAC

5 tmux panes open. Agents research, build, and review in parallel.

Two Ways to Use It

| | Local Mode | Hub Mode | |---|---|---| | Start | omt -d | omt hub start | | Interface | Terminal + tmux panes | Telegram / Slack (or terminal) | | Projects | One at a time | Multiple, always alive | | Remote | No | Yes, via channels from any device | | Best for | Active coding sessions | Background work, monitoring, multi-project |

Why Oh My Team?

| Without | With Oh My Team | |---------|----------------| | One AI, one project, one terminal | Multiple projects, always running | | Close laptop = work stops | Hub keeps sessions alive | | No visibility while away | Telegram/Slack shows progress + asks permission | | One agent does everything | 12 specialists work in parallel | | No quality gates | 5-agent review catches what you miss |

Installation

Let your LLM agent do it (recommended)

Copy this into Claude Code, Cursor, or any LLM agent:

Install and configure oh-my-team by following the instructions here:
https://raw.githubusercontent.com/erkandogan/oh-my-team/master/docs/install.md

The agent handles prerequisites, installation, and configuration. Humans fat-finger configs.

npm (manual)

brew install tmux                          # macOS (or apt install tmux on Linux)
npm i -g oh-my-team                        # install plugin + CLI
omt -d                                     # start

Claude Code Plugin

Inside a Claude Code session:

/plugin marketplace add erkandogan/oh-my-team
/plugin install oh-my-team

Then /reload-plugins to activate.

Requirements

• Claude Code v2.1.80+
• tmux — split-pane agent view and persistent sessions
• Bun — hub/channel features only (Telegram, Slack)

Full guide

See the Installation Guide for prerequisites, troubleshooting, and hub setup.

Uninstall

rm -rf ~/.oh-my-team ~/.local/bin/omt

Quick Start

# Start Oh My Team (launches inside tmux automatically)
omt

# With auto-approve permissions
omt -d          # shortcut for --dangerously-skip-permissions
omt --danger    # same thing

# Continue last session
omt -c

# All claude flags work
omt -d -c       # danger mode + continue

Your first team

1. Start: omt -d
2. Trigger team mode with the /oh-my-team:team command:

/oh-my-team:team Build a REST API with user authentication

Sisyphus creates the team, spawns specialized agents in tmux panes, and coordinates them through completion.

When to use /team vs direct prompts

• Use /oh-my-team:team when you want multiple agents working in parallel — research, implementation, review
• Use a direct prompt (no skill) for single-step tasks: "fix this typo", "explain how X works", "add error handling to this function"

Without /team, Claude Code will handle simple tasks directly — which is often faster. Use the skill when the task benefits from parallel specialists.

Agents

Oh My Team provides 12 specialized agents, each with a focused role and optimized system prompt.

Hub Layer

| Agent | Role | Model | |-------|------|-------| | Hub | Session manager. Manages multiple project sessions remotely via Telegram/Slack. Never writes code. | Sonnet |

Orchestration Layer

| Agent | Role | Model | |-------|------|-------| | Sisyphus | Team lead. Works directly on simple tasks, auto-escalates to planning/teams/review for complex work. | Opus | | Atlas | Plan conductor. Reads work plans, delegates to workers in parallel waves, verifies every result. | Sonnet |

Planning Layer

| Agent | Role | Model | |-------|------|-------| | Prometheus | Strategic planner. Interviews users, researches codebase, generates detailed work plans. Read-only. | Opus | | Metis | Gap analyzer. Catches hidden intentions, ambiguities, and scope creep risks before plan generation. | Opus | | Momus | Plan reviewer. Validates plans for executability. Approval-biased -- only rejects for true blockers. | Opus |

Worker Layer

| Agent | Role | Model | |-------|------|-------| | Hephaestus | Autonomous implementation worker. Full tool access. Explores patterns, implements end-to-end. | Opus | | Explorer | Fast codebase search. Finds files, patterns, and code structure. Multiple can run in parallel. | Haiku | | Librarian | Documentation and OSS research. Finds official docs, best practices, production patterns. | Haiku | | Oracle | Architecture consultant. Read-only. Pragmatic minimalism. Use for hard debugging and design decisions. | Opus |

Review Layer

| Agent | Role | Model | |-------|------|-------| | Reviewer | Code quality across 10 dimensions: correctness, patterns, naming, errors, types, performance, abstraction, testing, API design, tech debt. | Opus | | Security Auditor | 10-point security checklist: input validation, auth/authz, secrets, data exposure, dependencies, crypto, path traversal, error leakage. | Opus |

Skills

Skills are slash commands that trigger workflows.

Team skills (spawn agent teams)

| Skill | Purpose | |-------|---------| | /oh-my-team:team | Spawn an agent team with tmux panes for any task | | /oh-my-team:plan | Prometheus interview + Metis gap analysis + optional Momus review | | /oh-my-team:start-work | Atlas reads a plan and orchestrates execution with workers | | /oh-my-team:review-work | 5-agent parallel review: goals, QA, quality, security, context | | /oh-my-team:deep-debug | Multi-hypothesis parallel debugging with competing investigators |

Utility skills (no team, solo operation)

| Skill | Purpose | |-------|---------| | /oh-my-team:git-master | Atomic commit workflow with logical grouping | | /oh-my-team:ai-slop-remover | Detect and remove AI-generated code patterns | | /oh-my-team:frontend-ui-ux | Frontend development guidance (auto-loaded for UI work) |

Workflow: Plan, Execute, Review

1. /oh-my-team:plan "Add OAuth authentication"
   -> Prometheus interviews you, researches codebase, generates plan

2. /oh-my-team:start-work auth-plan
   -> Atlas reads plan, spawns workers, executes in parallel waves

3. /oh-my-team:review-work
   -> 5 agents review in parallel: goals, QA, quality, security, context

Workflow: Deep Debug

/oh-my-team:deep-debug "Login returns 500 after token refresh"
-> Spawns 3-5 investigators with competing hypotheses
-> They challenge each other's findings
-> Converge on the root cause

Architecture

+-------------------------------------------------+
|              Oh My Team Plugin                   |
+-------------------------------------------------+
|                                                  |
|  Hub            Hub agent (session manager)      |
|  Layer          Router + Bridge + Adapters       |
|                 Telegram | Slack                   |
|                                                  |
|  Orchestration  Sisyphus (lead)                  |
|  Layer          Atlas (conductor)                |
|                                                  |
|  Planning       Prometheus -> Metis ->           |
|  Layer          Momus (review)                   |
|                                                  |
|  Worker         Hephaestus (build)               |
|  Layer          Explorer (search)                |
|                 Librarian (research)             |
|                 Oracle (consult)                 |
|                                                  |
|  Review         Reviewer (quality)               |
|  Layer          Security Auditor (security)      |
|                                                  |
+-------------------------------------------------+
|  Team       plan | start-work | review-work      |
|  Skills     team | deep-debug                     |
+-------------------------------------------------+
|  Utility    git-master | ai-slop-remover          |
|  Skills     frontend-ui-ux                        |
+-------------------------------------------------+
|  Status    Agent name | Team | Members            |
|  Line      Context bar | Cost | Rate limits       |
+-------------------------------------------------+
             Claude Code Plugin System

How it works

Oh My Team is a Claude Code plugin with a lightweight channel system. The core is pure Markdown (agents + skills), the hub adds a small TypeScript bridge for remote control. It leverages Claude Code's native systems:

• Agents (agents/*.md) -- 12 specialized agents with model and tool configurations
• Skills (skills/*/SKILL.md) -- 8 slash commands that trigger workflows
• Channel system (channel/) -- Hub, router, bridge, and platform adapters for remote control
• Agent Teams -- Claude Code's experimental multi-session coordination
• Status Line -- Custom status bar showing active agents and teams

Hub — Remote Multi-Project Management

Control multiple projects from any device. The hub runs on your machine and connects to Telegram or Slack (Discord coming soon). Each project gets its own topic/thread — zero cross-talk, zero extra token cost.

Telegram Group: "Oh My Team Hub"
+-- General topic     <-- Hub: manage sessions
+-- Topic: frontend   <-- Bridge: direct to Claude (frontend)
+-- Topic: backend    <-- Bridge: direct to Claude (backend)
+-- Topic: mobile     <-- Bridge: direct to Claude (mobile)

How it works

• Hub session listens on General topic. Manages sessions via CLI.
• Each project gets its own Telegram topic + bridge (MCP channel server).
• You talk in a project's topic and the bridge forwards directly to that Claude session.
• Claude responds directly in the topic. Hub is never involved — zero extra token cost.

Setup (one-time)

Tell your LLM agent:

Set up Oh My Team hub by following:
https://raw.githubusercontent.com/erkandogan/oh-my-team/master/docs/hub-telegram.md

or for Slack:

Set up Oh My Team hub by following:
https://raw.githubusercontent.com/erkandogan/oh-my-team/master/docs/hub-slack.md

Or run the interactive wizard:

omt hub init     # choose platform, paste tokens, done
omt hub start    # starts router + hub session

Detailed guides: Telegram Setup | Slack Setup

Usage from Telegram

Talk to the hub in the General topic using natural language:

You: I want to work on the landing page project
Hub: Found ~/Desktop/projects/ohmyteam.cc — starting session...
Hub: Done! Switch to the "ohmyteam-cc" topic to talk to it.

You: what's running?
Hub: 2 active sessions:
  1. frontend — ~/projects/frontend (3h)
  2. ohmyteam-cc — ~/Desktop/projects/ohmyteam.cc (5m)

You: stop frontend
Hub: Stopped.

In a project topic, talk directly to that project's Claude session:

You: fix the login bug and add rate limiting
Claude: [works on it, replies when done]

You: run the tests
Claude: All 47 tests passing.

Permission prompts appear in the project topic — reply yes <code> or no <code>.

The hub finds projects by name — no need to type full paths.

Usage from terminal

omt hub add ~/projects/my-app   # add a session manually
omt hub list                     # list active sessions
omt hub attach my-app            # jump into a session
omt hub remove my-app            # stop a session
omt hub stop                     # stop everything

Supported platforms

| Platform | Status | Thread model | Setup guide | |----------|--------|-------------|-------------| | Telegram | Available | Forum Topics | Guide | | Slack | Available | Channel Threads (Socket Mode) | Guide | | Discord | Coming soon | Channels or Threads | -- |

Adding a new platform = one adapter file (~150 lines) implementing the ChannelAdapter interface. Bridge and router are platform-agnostic.

Requirements

• Bun runtime (for the channel server)
• Claude Code v2.1.80+ with claude.ai login
• tmux (for persistent sessions)

Note: Channels are a Claude Code research preview feature. Custom channels require the --dangerously-load-development-channels flag.

Status Line

Oh My Team installs a custom status line at the bottom of Claude Code:

@sisyphus [my-team] 4 agents: explorer-1 librarian-1 builder-1 builder-2
|||||||________ 45% | $1.23 | 5m30s | 5h:18% 7d:5%

Shows: active agent, team name, teammate count, context usage (color-coded), session cost, duration, and rate limits.

Project Structure

oh-my-team/
+-- docs/
|   +-- install.md               # LLM-friendly installation guide
|   +-- hub-telegram.md          # Telegram hub setup
|   +-- hub-slack.md             # Slack hub setup
+-- .claude-plugin/
|   +-- plugin.json              # Plugin manifest
+-- agents/                      # 12 specialized agents
|   +-- sisyphus.md              # Team lead / orchestrator
|   +-- hub.md                   # Session manager (remote control)
|   +-- prometheus.md            # Strategic planner
|   +-- atlas.md                 # Plan conductor
|   +-- oracle.md                # Architecture consultant
|   +-- hephaestus.md            # Implementation worker
|   +-- explorer.md              # Codebase search
|   +-- librarian.md             # Documentation research
|   +-- reviewer.md              # Code quality review
|   +-- security-auditor.md      # Security review
|   +-- metis.md                 # Gap analyzer
|   +-- momus.md                 # Plan reviewer
+-- skills/                      # 8 slash commands
|   +-- team/SKILL.md            # Force team mode
|   +-- plan/SKILL.md            # Planning workflow
|   +-- start-work/SKILL.md      # Execution workflow
|   +-- review-work/SKILL.md     # 5-agent review gate
|   +-- deep-debug/SKILL.md      # Multi-hypothesis debugging
|   +-- git-master/SKILL.md      # Commit workflow (utility)
|   +-- ai-slop-remover/SKILL.md # Code cleanup (utility)
|   +-- frontend-ui-ux/SKILL.md  # Frontend guidance (auto-loaded)
+-- channel/                     # Hub channel system
|   +-- bridge.ts                # MCP channel server (per-session)
|   +-- router.ts                # Session registry + message broker
|   +-- adapters/
|   |   +-- types.ts             # ChannelAdapter interface
|   |   +-- telegram.ts          # Telegram Topics adapter
|   |   +-- slack.ts             # Slack Threads adapter (Socket Mode)
|   |   +-- index.ts             # Adapter registry
|   +-- slack-manifest.yml       # One-click Slack App manifest
|   +-- bridge.mcp.json          # MCP config for bridge
|   +-- package.json             # Channel dependencies
+-- bin/
|   +-- omt                      # CLI (sessions + hub management)
+-- hooks/hooks.json             # Session lifecycle hooks
+-- settings.json                # Activates Sisyphus as default
+-- CLAUDE.md                    # Project instructions
+-- install.sh                   # One-command installer

Contributing

Contributions are welcome! See CONTRIBUTING.md for guidelines.

Adding a new agent

Create agents/your-agent.md with frontmatter:

---
name: your-agent
description: "What this agent does and when to use it"
model: opus  # or sonnet, haiku
tools: [Read, Grep, Glob, Bash]  # tool restrictions
---

Your agent's system prompt here.

Adding a new skill

Create skills/your-skill/SKILL.md with frontmatter:

---
name: your-skill
description: "What this skill does"
argument-hint: "[arguments]"
---

Instructions for the skill.

Adding a new channel adapter

Create channel/adapters/your-platform.ts implementing the ChannelAdapter interface from types.ts. Register it in adapters/index.ts and add a case in router.ts. See telegram.ts for a complete reference implementation.

License

MIT -- Use it, fork it, make it yours.

Credits

Inspired by Oh My OpenCode by YeonGyu Kim. Built for Claude Code by Anthropic.