Back to directory

Pi Extensions

Pi Extensions is a monorepo of independently installable extensions for the Pi coding agent, covering coding and delegation, research, browser automation, workflow management, observability and terminal ergonomics.

Pi Extensions

Pi Extensions collects a set of separately installable extensions and reusable extension libraries for the Pi coding agent, published under one npm scope so a user installs only the ones they want. Examples include a side-question tool that answers without disturbing the main conversation, an account switcher for named OAuth subscriptions, a tool to keep the machine awake during long runs, higher-quality compaction for Codex sessions, response timestamps, matching a terminal's Starship prompt, syncing configuration across devices through S3, and checking Codex usage limits without opening Codex.

A coding and delegation group includes an extension for OpenAI Codex Remote Compaction with manual controls and a safe fallback, an experimental context-rollover extension with bounded branch-history recall, a file browser for attaching exact lines or git diff hunks with provenance to a prompt, and a language-server extension providing diagnostics and code actions across many languages.

Extensions run with the user's full permissions, so the project recommends reviewing any third-party extension before installing it; each is written in TypeScript for Node.js.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.