pi-coding-agent

Pi coding-agent extension: adds an apply_patch tool (V4A diffs) so GPT-5.2/Codex can propose structured patches that Pi applies.

Pi extension that generates a per-turn system-prompt augmentation + senior persona via a nested Pi call (openai-codex), optimized for design-first full-stack craftsmanship.

Fast local semantic search across AI agent skill registries with quality filtering. Built in Rust.

Reusable skills for AI coding agents (pi, Claude Code, Agent Skills standard)

🧠 pi-readcache — Replay-aware read caching for Pi (lower token usage, compaction-safe correctness).

📚 Batch file reads for Pi via read_many with adaptive packing, safe heredoc framing, and output-budget awareness

Universal MCP tools extension for pi coding agent — connect to filesystem, GitHub, databases, and more

GNU Readline-style Ctrl+R reverse search extension for pi prompt and bash history.

🧜‍♀️ Pi extension that renders Mermaid diagrams as ASCII in the TUI, with width-aware output and safe handling for larger diagrams.

AI agents for genomics

Exa-powered source discovery and search workflows for Pi.

Pi package that adds a /r command with cached ripgrep-backed file completions

A Pi extension that blocks prompts containing polite filler before they reach the LLM.

pi-lab lets you run tool-based A/B tests in various modes

Exa-powered source discovery and search workflows for Pi.

A 100% local, containerized AI coding agent powered by pi coding agent and llama.cpp. Run private LLMs on CPU or NVIDIA GPU without external API dependencies.

📱 Control pi coding agent via iMessage — text prompts from your iPhone, get AI responses back. Pi package + standalone daemon.

Token Killer for Pi — reduce LLM token consumption by 60-90% on common dev commands

A small Pi coding agent extension to respect explicit `reasoning-effort` from the loaded `SKILL.md`.