oh-my-openagent

Oh My OpenAgent (OmO) is an open-source AI coding agent toolkit designed to streamline software development through multi-model orchestration and automated workflow execution. It is available in three editions to fit different user environments: Ultimate edition for OpenCode, Light edition for OpenAI Codex CLI, and a standalone Senpi beta edition. The Ultimate edition is a full OpenCode plugin that includes 11 specialized agents, over 54 lifecycle hooks, 4 built-in MCP servers (websearch, context7, grep_app, lsp), Team Mode, the /goal persistent objective feature, ultrawork mode, and optional hashline edit functionality. The Light edition is a portable plugin for Codex CLI that includes core components like rule injection, comment checking, git-bash integration, LSP support, ultrawork, ulw-loop, ulw-execute continuation, telemetry, and teammode, along with plugin-scoped MCPs for grep_app, context7, git_bash, and lsp, plus the shared ast-grep skill. The standalone Senpi beta edition is installed via npm i -g omo-ai@beta and ships a native omo command with the OMO extension built in, no host application required. Core capabilities include the ultrawork (ulw) keyword-triggered mode that activates all agents to complete tasks without manual intervention, Discipline Agents for parallel task orchestration, Team Mode that supports up to 8 parallel specialized agents with real-time tmux terminal visualization, and specialized modes like hyperplan (with 5 hostile critic agents for planning) and security-research (with 3 hunter and 2 PoC engineer agents). It integrates LSP for IDE-level code diagnostics, navigation, and workspace rename support, and AST-Grep for pattern-aware code search and rewriting across 25 programming languages. Additional features include hash-anchored edit tools to eliminate stale line errors, a comment checker to maintain clean code comments, automatic rules injection from AGENTS.md and .omo/rules/ files, persistent goal tracking via /goal, a todo enforcer to keep agents on task, compatibility with Claude Code hooks, commands, skills, MCPs and plugins, skill-embedded MCPs to reduce context bloat, an ultrawork planner for interview-mode strategic planning before execution, and /init-deep for auto-generating hierarchical AGENTS.md files across projects. The toolkit supports multiple LLM providers including Anthropic, Gemini, Kimi, GPT, and others, with no lock-in to a single model or provider. It includes optional anonymous telemetry that can be disabled via configuration or environment variables, and is positioned as an open alternative to walled-garden coding tools like Claude Code and Cursor. The project is actively maintained, with a public roadmap for supporting additional agent harnesses beyond OpenCode and Codex CLI.