Pi is an open-source AI agent toolkit centered on a unified multi-provider LLM API, an agent runtime, a terminal UI, and an interactive coding agent CLI.
Pi is a monorepo for an agent harness project, including an interactive coding agent and several supporting packages. The README identifies packages for a unified LLM API, agent runtime with tool calling and state management, a terminal UI library, and telemetry contracts and tests. It is presented as the home of the Pi agent harness and points users to documentation and demos on pi.dev.
It addresses the need to build and run AI agents using multiple LLM providers without forcing each layer to be custom-built from scratch. The project also aims to support interactive coding workflows, terminal-based interaction, and reusable agent runtime components in one ecosystem.
At a high level, Pi combines a provider-agnostic LLM API with an agent runtime that manages tool use and state, then exposes that through an interactive coding agent CLI and a terminal UI library. The README also mentions vendor-neutral telemetry contracts, reference adapters, conformance tests, and typed schemas, suggesting a structured way to observe and validate integrations. Beyond that, the README does not describe the internal execution model in detail.
It is gaining attention because the repository is directly focused on AI agent tooling, a fast-moving area, and it reports strong recent momentum with 87,680 stars and 18,096 gained this month. The README also emphasizes practical developer-facing pieces such as a coding agent CLI, documentation, demos, and release/build tooling, which likely makes it easier for users to try.
The README names pi-chat as a related project for Slack/chat automation and workflows. More broadly, the package layout suggests Pi overlaps with other agent frameworks, LLM abstraction layers, terminal UI toolkits, and coding-agent CLIs, but the README does not compare it directly against specific competitors.
AI-explained · grounded in each repo's README