Pi vs mini-swe-agent
AI coding agents for development workflows compared
pi and mini-swe-agent both provide command-line AI agents for software engineering tasks. pi is a TypeScript toolkit built around a multi-provider API and extensive customization, while mini-swe-agent is a compact Python agent that uses bash and a linear message history.

Pi: Build and Customize a Coding Agent
Pi is a minimal, extensible coding-agent harness built with TypeScript and its pi-ai unified LLM API. Run it as a CLI on macOS, Linux, or Windows, or adapt its components and extensions to your own workflows.

mini-swe-agent: Automate Software Engineering Tasks with an AI Agent
mini-swe-agent is a compact Python agent that uses a language model and bash to tackle software engineering tasks, including GitHub issues. Its simple control flow suits developers who want a hackable CLI or a baseline for agent research.
| Pi | mini-swe-agent | |
|---|---|---|
| Language | TypeScript | Python |
| License | MIT | MIT |
| Stars | 117k | 8.2k |
| Forks | 15k | 1.1k |
| Last analyzed | Oct 23, 2026 | Oct 3, 2026 |
Key differences
- pi is written in TypeScript and includes an agent runtime, terminal UI, unified LLM API, and SDK; mini-swe-agent is written in Python and offers a CLI and Python bindings.
- pi supports workflow customization through extensions, skills, prompt templates, themes, and packages; mini-swe-agent emphasizes a small scaffold and bash as its agent interface.
- pi offers print and JSON modes, RPC, and supporting libraries for integration and automation; mini-swe-agent includes batch inference for SWE-bench and a trajectory browser.
- mini-swe-agent documents support for local, container, and other execution environments; pi runs with the launching user's permissions and recommends sandboxing or containers when stronger isolation is needed.
- Both projects use the MIT license. The supplied project facts list pi with 112k stars and 14.2k forks, and mini-swe-agent with 8.2k stars and 1.1k forks.
Choose Pi if you…
- want a TypeScript toolkit with a unified API for multiple LLM providers.
- need extensions and packages to adapt agent tools, prompts, or interface to a workflow.
- plan to build TypeScript applications around an agent or automate runs with JSON output or RPC.
Choose mini-swe-agent if you…
- prefer a compact Python agent that uses bash rather than a collection of custom tools.
- want a linear, inspectable history for adapting or evaluating agent behavior.
- need documented options for running tasks in local or sandboxed environments.
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.