Project overview
Pi is an extensible AI agent toolkit. It provides both a terminal-based coding assistant and TypeScript libraries for building applications, including a unified multi-provider model API, tool calling, state management, and terminal UI capabilities.
This is a software repository, not a resource directory. Although its listing topic is video generation, the supplied README does not describe video generation capabilities, so it should not be presented as a video generation tool.
Use cases
Pi suits users who want an AI coding assistant working in a project directory, as well as developers integrating agent capabilities into applications or automated workflows.
It can be used interactively, automated through print or JSON mode, controlled from another program through RPC, or integrated into applications through its TypeScript SDK. Workflows can be customized with extensions, skills, prompt templates, and themes, then bundled as Pi packages and shared through npm or Git.
What it does and problems it solves
Pi provides a unified interface for multiple model providers, alongside tool calling and state management for agents carrying out tasks. These libraries let developers reuse foundational agent capabilities. Its terminal UI library and coding assistant support command-line interaction with an agent.
The defaults do not include sub-agents or plan mode. Pi also has no built-in permission system restricting filesystem, process, network, or credential access; it uses the permissions of the user and process that launch it. If isolation is needed, consult the original repository's containerization documentation for the Gondolin extension, Docker, or OpenShell options.
How to use
1. Install Pi using the instructions below and prepare the project directory where you want it to work.
2. Open a terminal and enter `cd /path/to/project`, replacing `/path/to/project` with your actual project directory path.
3. Enter `pi` to start the interactive assistant.
4. Enter `/login` inside Pi to connect a subscription or API key for a built-in AI provider. The supplied material does not provide pricing information for these services.
5. Describe the task you want Pi to perform. For automation, RPC control, SDK integration, or custom extensions, consult the corresponding documentation in the original repository; the supplied README does not include complete operating parameters for these workflows.
Installation
1. Prepare Node.js 22.19 or newer, which is the runtime requirement stated in the README. The supplied material does not give separate instructions for installing Node.js.
2. For the npm installation option, run `npm install -g --ignore-scripts @earendil-works/pi-coding-agent` in a terminal. The README states that normal npm installation does not require dependency lifecycle scripts; this option does not pin all transitive dependency versions.
3. Alternatively, obtain the official installer entry point from the original repository below this page and follow its instructions for your system. The README provides installers for macOS, Linux, and Windows, stating that they can install Node.js if needed and pin dependency versions. Installer-based installations can be updated with `pi update`.
4. After installation, follow the usage steps above to enter your project directory and run `pi`. This listing only provides the repository link; runnable packages and core workflows have not been validated, so successful operation after installation is not guaranteed here.
Source and usage notes
- Original repository: https://github.com/earendil-works/pi
- Source revision: 58c0b02ce552adab76014e7ed712ab7981359aa8
This listing links to the upstream repository. It does not provide a verified download package, and the complete workflow has not been independently validated.
This description was organized and translated with AI from public upstream material. Refer to the original repository for capabilities, setup requirements, costs, and license terms. 
