The can1357/oh-my-pi repository is an advanced AI coding agent designed for terminal environments, offering a comprehensive suite of tools and integrations that transform the terminal into a powerful, IDE-like coding assistant. Built primarily in TypeScript, with a Rust core, oh-my-pi (OMP) is a fork of Mario Zechner’s Pi project, significantly expanding its capabilities and providing a complete, out-of-the-box solution for developers seeking AI-powered assistance in their coding workflows.
OMP’s primary purpose is to streamline and enhance coding productivity by leveraging large language models (LLMs) from over 40 providers, including OpenAI, Anthropic, Gemini, Grok, and MiniMax. The agent is optimized for real-world use, continuously tuned, and features more than 32 built-in tools, 13 LSP (Language Server Protocol) operations, and 27 DAP (Debug Adapter Protocol) operations. Its Rust core comprises approximately 27,000 lines, ensuring high performance and native integration across macOS, Linux, and Windows without reliance on external binaries or shell scripts.
A standout feature of OMP is its hash-anchored editing system, which enables precise, content-based edits. By referencing content hashes rather than line numbers, the agent avoids common pitfalls such as whitespace conflicts and ensures patch integrity, especially when files change or anchors diverge. This approach reduces token usage and increases reliability, particularly with advanced models like Grok 4 Fast.
OMP integrates deeply with LSP, allowing the agent to perform IDE-grade operations such as renaming symbols, updating re-exports, and managing aliased imports. The agent’s knowledge mirrors that of modern IDEs, ensuring that code refactoring and navigation are accurate and context-aware. Debugging is equally robust, with native support for lldb, dlv, and debugpy, enabling the agent to attach to processes, inspect frames, and evaluate expressions across languages like C, Go, and Python.
The tool harness is highly optimized, featuring persistent Python and Bun workers that can call back into the agent’s own tools via a loopback bridge. This allows for seamless execution of code, data analysis, and visualization within the terminal, without leaving the session. OMP’s web_search tool chains multiple ranked providers, enabling the agent to read and summarize content from sources like arxiv PDFs, GitHub pages, and Stack Overflow threads, returning structured markdown with anchors for easy citation and navigation.
OMP supports advanced workflows through first-class subagents, which can split tasks across isolated worktrees and return schema-validated results. This eliminates the need for parsing prose or resolving merge conflicts, making collaborative and parallel work efficient. The agent also offers prioritized code reviews, spawning dedicated reviewer subagents that rank issues by severity and confidence, providing clear verdicts on whether changes are ready to ship.
Memory management is another key feature, with the agent retaining and recalling facts about the codebase between sessions. This project-scoped memory ensures continuity and context, allowing the agent to build a mental model of the repository over time. OMP’s interface is unified, treating GitHub as just another filesystem, simplifying interactions with PRs and issues.
Installation is straightforward, with support for Bun, PowerShell, and pinned versions via mise. Shell completions are dynamically generated for bash, zsh, and fish, ensuring that CLI commands and flags are always up-to-date. Overall, oh-my-pi delivers a highly capable, native, and extensible AI coding agent for terminal users, bridging the gap between traditional IDEs and modern AI-powered development tools.
