DISCOVER THE FUTURE OF AI AGENTSarrow_forward

oh-my-agent

calendar_todayAdded Apr 24, 2026
categoryAgent & Tooling
codeOpen Source
TypeScriptWorkflow AutomationMulti-Agent SystemModel Context ProtocolAI AgentsAgent FrameworkCLIBunAgent & ToolingDeveloper Tools & CodingAutomation, Workflow & RPAProtocol, API & Integration

A portable multi-agent harness for `.agents`-based skills, workflows, and standards-aware agent teams across multiple AI IDEs and CLI tools.

oh-my-agent is a portable multi-agent orchestration framework built around the .agents/ standard directory. Its core design treats .agents/ as the single source of truth (SSOT), version-controlled alongside project code. The framework includes 21 specialized Agents, each mapping to a real engineering team role (frontend, backend, architecture, QA, PM, DB, mobile, DevOps, debugging, design, etc.), with independent toolchains, checklists, and domain knowledge.

Multi-Agent Orchestration#

  • 21 role-specific Agents, each with independent tools and checklists
  • Parallel execution: oma agent:parallel spawns multiple sub-agents concurrently, coordinated via MCP memory
  • Per-Agent model configuration: specify different models and effort levels per role in oma-config.yaml

Cross-IDE/CLI Portability#

  • Native sub-agent generation: auto-generates .claude/agents/*.md, .codex/agents/*.toml, .gemini/agents/*.md from .agents/agents/
  • Same-vendor runtimes use native sub-agent paths; cross-vendor scenarios fall back to oma agent:spawn
  • Supported runtimes: Claude Code, Gemini CLI, Codex CLI, Antigravity IDE, Cursor, OpenCode
  • Runtime profiles: claude-only, codex-only, gemini-only, antigravity, qwen-only

Skill Management#

  • Two-layer progressive loading: Layer 1 SKILL.md (~800 bytes, always loaded); Layer 2 resources/ (loaded on demand), reportedly saving ~75% tokens
  • Agent Package Manager (APM) integration: one-click deployment of 22 skills to all detected runtimes
  • 22 skill directories (21 agents + _shared)

Workflow Automation#

  • Auto workflow detection: UserPromptSubmit Hook scans natural language keywords (11 languages supported) to auto-activate workflows
  • 16 predefined workflows: brainstorm, architecture, design, plan, work, orchestrate, ultrawork, review, debug, scm, etc.
  • Quality gates: Charter pre-checks, multi-stage quality gates, OWASP security audits, WCAG accessibility reviews

Developer Experience#

  • HUD status bar: [OMA] status indicator in Claude Code (model name, context usage, workflow state)
  • Natural language interaction: describe needs directly, auto-routed to the corresponding Agent
  • i18n support: docs and keyword detection for Korean, Chinese, Japanese, French, Spanish, German, Portuguese, Russian, Dutch, Polish
  • Post-install presets: All / Fullstack / Frontend / Backend / Mobile / DevOps

Architecture & Implementation#

  • Monorepo structure: Bun workspaces with cli/ (CLI tool) and web/ (docs site, Docusaurus)
  • SSOT design: .agents/ is the single source of truth; .claude/, .codex/, .gemini/ are IDE integration layers only
  • Hook system: UserPromptSubmit (keyword detection), PreToolUse (tool call interception), Stop (persistent mode)
  • MCP protocol for inter-Agent memory sharing and Serena symbol tracing
  • Build toolchain: Biome, Husky, commitlint, mise, TypeScript, release-please
  • Primary languages: TypeScript (90.9%), Shell (5.5%), Rust (1.3%), etc.

Installation#

# One-click install (macOS/Linux)
curl -fsSL https://raw.githubusercontent.com/first-fluke/oh-my-agent/main/cli/install.sh | bash

# Manual install
bunx oh-my-agent@latest

# Global install
bun install --global oh-my-agent
# or
brew install oh-my-agent

# Via APM
apm install first-fluke/oh-my-agent

Core CLI Commands#

oma doctor                    # Health check
oma dashboard                 # Real-time Agent monitoring
oma link                      # Regenerate IDE configs from .agents/
oma agent:spawn backend "Build auth API" session-01
oma agent:parallel -i backend:"Auth API" frontend:"Login form"

Typical Use Cases#

  • Full-stack app development with multi-role collaboration (PM → Architecture → Frontend/Backend → DB → QA)
  • Architecture decision analysis (ADR/ATAM/CBAM-style trade-offs)
  • Code quality & security auditing (OWASP Top 10, WCAG 2.1 AA)
  • Infrastructure as Code (multi-cloud Terraform, OPA/Sentinel policies)
  • Debugging & root cause analysis, document conversion, multilingual translation

Current version: v5.18.2 (CLI) / web v1.1.3 | Repo activity: 1,377+ commits, 229+ releases | Platforms: macOS, Linux (official); Windows (manual bun/uv install, compatibility unverified)

Related Projects

View All arrow_forward

STAY UPDATED

Get the latest AI tools and trends delivered straight to your inbox. No spam, just intelligence.

rocket_launch