pi

Overview

Pi is an AI agent toolkit that provides a unified LLM API, an agent loop, a terminal UI, and a coding agent CLI. It includes multiple packages for building agents, telemetry, durable runtimes, and more.

Key Features

• Unified multi-provider LLM API supporting OpenAI, Anthropic, Google, and others
• Agent runtime with tool calling and state management
• Interactive coding agent CLI
• Terminal UI library with differential rendering
• Durable conversation, task, and document runtime

Use Cases

• Building custom AI agents with tool calling and state management
• Using the interactive coding agent for development tasks
• Creating terminal-based AI applications
• Implementing durable conversation and task systems

Getting Started

• For development, run `npm install --ignore-scripts` to install dependencies, then `npm run build` to build all packages. Use `./test.sh` to run tests, and `./pi-test.sh` to run pi from sources.

Deployment & Requirements

• Pi can be containerized or sandboxed for stronger boundaries. See the containerization documentation for patterns using Gondolin, Docker, or OpenShell. Standalone binaries can be built from release source archives using the provided build script.

Before You Adopt

• License: MIT. Review its terms before using, modifying, or distributing the project.
• Pi does not include a built-in permission system; it runs with the permissions of the launching user and process.
• New issues and PRs from new contributors are auto-closed by default, with maintainers reviewing them daily.