oh-my-pi

// Overview

oh-my-pi — `omp` on the command line — is a coding agent for the terminal, forked from Mario Zechner's Pi and extended into a batteries-included assistant. Its stated premise is "a coding agent with the IDE wired in": rather than editing text blind, the agent talks to a language server, so renames and refactors happen at workspace level with the same symbol information a human editor would have.

The debugging and execution story follows the same idea. omp drives real debuggers through DAP — lldb, dlv and debugpy — and keeps persistent Python and JavaScript sessions alive between tool calls, so state carries across steps instead of every snippet starting from scratch. It ships 31 built-in tools, subagents that run work in parallel and return schema-validated results, a memory system for durable facts and learned skills, and web search across 23 providers with structured extraction.

It is deliberately provider-neutral: 60+ LLM providers are supported, spanning frontier APIs (Anthropic, OpenAI, Google Gemini, xAI, DeepSeek, Mistral, Groq), coding plans routed through Cursor, GitHub Copilot and Devin, and local runtimes via Ollama, LM Studio and llama.cpp. The core is roughly 80,000 lines of Rust with a TypeScript/Bun surface, ripgrep, bash and coreutils are linked in-process, and live collaboration is available through shareable links and QR codes. The project is MIT-licensed; vendored components such as brush-shell keep their upstream licences.

// What it does

• LSP integration so edits and refactors are symbol-aware across the whole workspace
• Debugger support through DAP with lldb, dlv and debugpy
• Persistent Python and JavaScript execution environments that keep state across tool calls
• 60+ LLM providers: frontier APIs, coding plans (Cursor, GitHub Copilot, Devin) and local models via Ollama, LM Studio and llama.cpp
• Subagents that run in parallel and return schema-validated results
• Memory system for durable facts and learned skills, plus 31 built-in tools
• Web search across 23 providers with structured extraction
• Live collaboration through shareable links and QR codes

// Getting started

omp installs as a single command-line tool. Pick whichever of the project's install paths fits your machine, then run it inside a repository.

curl -fsSL https://omp.sh/install | sh
# or
brew install can1357/tap/omp

bun install -g @oh-my-pi/pi-coding-agent

irm https://omp.sh/install.ps1 | iex
# Nix
nix run github:can1357/oh-my-pi

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

// When to use it

• Reach for it when you want an agent whose edits are language-server accurate rather than text-substitution guesses
• Use it when a task needs real debugging — setting breakpoints and stepping through with lldb, dlv or debugpy — not just re-running a script
• Good for data-ish work where a persistent Python or JavaScript session between steps saves recomputing state
• Useful when you want one agent front-end across frontier APIs, existing coding plans and local models

// How oh-my-pi compares

oh-my-pi alongside other open-source autonomous coding agents tools AI/TLDR tracks, ranked by GitHub stars.