Pi vs harness-sdk

Coding-agent harnesses and agent-building SDKs compared

Pi is a TypeScript coding-agent harness designed for project-based use and customization. harness-sdk provides Python and TypeScript SDKs for building agents, with assembled harnesses and controls for the agent loop; the main difference is Pi’s coding-agent focus versus harness-sdk’s broader agent-building scope.

Piharness-sdk
LanguageTypeScriptPython
LicenseMITApache-2.0
Stars117k9k
Forks15k1.4k
Last analyzedOct 23, 2026Oct 29, 2026

Key differences

  • Pi is a TypeScript coding-agent CLI and set of reusable libraries; harness-sdk offers Python and TypeScript SDKs, assembled harnesses, a CLI, and an MCP server.
  • Pi supports interactive use, scripting, RPC control, and embedding through a TypeScript SDK; harness-sdk focuses on building and running agents with control over their in-process loop.
  • Pi provides extensions, skills, prompt templates, themes, and packages for customization; harness-sdk includes assembled defaults for models, tools, memory, sessions, and context management.
  • Pi is licensed under MIT; harness-sdk is licensed under Apache-2.0.
  • Pi runs on macOS, Linux, and Windows, while harness-sdk has separate Python and TypeScript toolchain requirements, including Python 3.10 or later and Node.js 22 or later for its TypeScript SDK.
  • Pi explicitly leaves out built-in sub-agents and plan mode; harness-sdk lists multi-agent patterns and lifecycle controls such as turn limits and token budgets.

Choose Pi if you…

  • want a customizable coding-agent CLI for work in a project directory.
  • need print or JSON modes, RPC control, or a TypeScript SDK for integrations.
  • prefer to add workflow features such as sub-agents or plan mode through extensions or packages.
Read the Pi analysis →

Choose harness-sdk if you…

  • want to build agents in Python or TypeScript around an in-process agent loop.
  • need assembled harness defaults or controls such as turn limits, token budgets, and cancellation.
  • plan to use features such as MCP connections, guardrails, tracing, or evaluation support.
Read the harness-sdk analysis →

Compare the numbers side by side, or add more projects

This comparison is generated with AI from the OSRepos analyses of both projects. Always check each project's repository and documentation before choosing.