What problem it solves

Useful software engineering agents can require complex control flows, specialized tools, extensive configuration, and large dependency footprints. This project provides a simple, hackable agent that can work through the shell and run in local or sandboxed environments.

Key features

• Shell-only agent — The agent interacts with its environment through bash without requiring model-specific tool calls. This keeps its interface small and makes it suitable for sandboxed execution.

  Source: README

• Simple, inspectable control flow — The agent keeps a completely linear history in which each step appends to the messages sent to the model. This structure is designed to make behavior easier to understand, debug, and fine-tune.

  Source: README

• Flexible deployment and model support — The agent supports local environments and several container and sandbox systems. It also supports models through integrations including litellm, openrouter, and portkey.

  Source: README

• Command-line and Python interfaces — Users can run the agent from the command line, inspect trajectories, perform batch inference, or invoke it through Python bindings.

  Source: README

Quick start

Copied verbatim from the project's README.

pip install mini-swe-agent
mini  # run the CLI

agent = DefaultAgent(
    LitellmModel(model_name=...),
    LocalEnvironment(),
)
agent.run("Write a sudoku game")

Who it may suit

Inference: This project may suit developers who want a lightweight command-line coding agent, researchers building agent baselines, and practitioners running fine-tuning or reinforcement-learning experiments without a complex agent scaffold.

Topics to discuss

• What capabilities can be added to a minimal shell-based agent while preserving its simplicity?

• How should independent subprocess execution be tested across local and sandboxed environments?

• Which model and endpoint configurations are most useful for daily software engineering workflows?

Written by ReadmeX AI from the project's README, source and public GitHub data; it may contain mistakes.

Read the original README https://github.com/SWE-agent/mini-swe-agent#readme

ReadmeX opened this community for a trending project; it is not run by the project. Maintainers can claim it from the community page.