Oh My Pi FAQ

Oh My Pi (omp) is an AI coding agent with the IDE wired in. It's a fork of Pi by Mario Zechner, extended with a batteries-included coding workflow. It features hash-anchored edits, optimized tool harness, LSP integration, Python execution, browser automation, subagents, and much more. Built with TypeScript and Rust, it provides a complete coding assistant that works directly in your terminal.

Unlike other AI coding assistants, Oh My Pi:

• Uses hash-anchored edits instead of line numbers, reducing token usage by 61%
• Integrates LSP directly into every write operation
• Supports real debugging with lldb/dlv/debugpy
• Includes time-traveling stream rules for course correction
• Offers first-class subagents for parallel processing
• Runs persistent Python and JavaScript kernels
• Has native performance with no fork-exec on hot paths

Oh My Pi supports:

• macOS (Intel and Apple Silicon)
• Linux (x86 and ARM)
• Windows (via PowerShell installer)

It runs on Bun runtime (recommended) or Node.js. The same binary works across all platforms without WSL bridge.

Installation is simple: Or using Bun (recommended): Windows PowerShell: Homebrew:

Oh My Pi supports 40+ providers including:

• Frontier APIs: Anthropic, OpenAI, Google Gemini, xAI, Mistral, Groq
• Coding plans: Cursor, GitHub Copilot, GitLab Duo, Kimi Code, MiniMax
• Self-hosted: Ollama, LM Studio, llama.cpp, vLLM, LiteLLM
• Custom providers with OpenAI-compatible endpoints

Models can be switched mid-session with Ctrl+P or the /model command.

Yes! Oh My Pi integrates directly with editors through the Agent Client Protocol (ACP). When running inside Zed, it reads the buffer you're looking at and writes through the editor's save path. It supports permission prompts for destructive tools and maintains synchronization without any bridge or plugin needed. The same agent works in the terminal and in your IDE.

Oh My Pi includes 32 built-in tools:

• Files & Search: read, write, edit, ast_edit, ast_grep, search, find
• Runtime: bash, eval (Python/JS), ssh
• Code Intelligence: lsp, debug
• Coordination: task, irc, todo, job, ask
• Outside the Box: browser, web_search, github, generate_image, inspect_image
• Memory & State: checkpoint, rewind, retain, recall, reflect

Use /collab to share live sessions:

• Share read-write access to pair program
• Share read-only links for team reviews
• Generate QR codes for easy joining
• Frames are sealed client-side (keys never sent to relay)
• Teammates can join with `omp join` or browser link

Perfect for remote pair programming and code reviews.

Oh My Pi is open-source under the MIT License. You can use it for free with any supported provider. Some providers (like GitHub Copilot, Cursor, Kimi Code) require their own subscriptions. For local models (Ollama, LM Studio), there are no additional costs beyond your hardware requirements.

Contributions are welcome! The project follows a vouch system:

• Open a Discussion to request vouching
• Once vouched, you can submit PRs
• Code must follow existing patterns and style
• Include tests for new functionality
• Update documentation as needed

See CONTRIBUTING.md for detailed guidelines. We also welcome bug reports and feature requests via GitHub Issues.

Get help through:

• Discord: Join our community at discord.gg/4NMW9cdXZa
• GitHub Issues: Report bugs or request features
• Documentation: Full docs at omp.sh/docs
• Changelog: Track updates in packages/coding-agent/CHANGELOG.md

The community is active and helpful for both beginners and experienced users.

Oh My Pi is built with:

• Core: ~55,000 lines of Rust for in-process tools
• Surface: TypeScript for agent logic and TUI
• Runtime: Bun or Node.js with N-API native addons
• Storage: SQLite for memory (Hindsight)
• Protocol: Agent Client Protocol for editor integration

The architecture is designed for native performance with minimal overhead.

Yes! Oh My Pi is released under the MIT License, which allows commercial use. You can integrate it into commercial products, use it for commercial development, and modify it according to your needs. The only requirement is to include the original license and copyright notice in your distributions.

Minimum requirements:

• OS: macOS 10.14+, Linux (Ubuntu 18.04+), Windows 10+
• RAM: 4GB minimum, 8GB recommended
• CPU: x86_64 or ARM64 architecture
• Storage: 100MB for installation
• Internet: Required for cloud model providers

For local models, more RAM and GPU acceleration (optional) will improve performance.

Oh My Pi includes Hindsight, a memory system that:

• Caches facts during sessions with `retain`
• Recalls memories with `recall`
• Synthesizes answers over the bank with `reflect`
• Compresses sessions into mental models
• Loads project-scoped memory automatically

This helps the agent remember your codebase between sessions and improve over time.