trae-agent vs mini-swe-agent
Python agents for software engineering compared
trae-agent and mini-swe-agent are Python CLI agents that use language models to carry out software engineering tasks. trae-agent offers configurable tools and providers, while mini-swe-agent centers on bash and a compact, linear workflow.

trae-agent: LLM Agent for Software Engineering
Trae Agent is a Python CLI agent that uses configurable LLM providers to carry out general-purpose software engineering tasks. Its modular tools, interactive mode, and recorded trajectories suit developers and researchers who want to run or study agent 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.
| trae-agent | mini-swe-agent | |
|---|---|---|
| Language | Python | Python |
| License | MIT | MIT |
| Stars | 12.2k | 8.2k |
| Forks | 1.4k | 1.1k |
| Last analyzed | Oct 23, 2026 | Oct 3, 2026 |
Key differences
- trae-agent provides tools for file editing, bash execution, sequential thinking, and task completion; mini-swe-agent uses bash as its agent interface.
- trae-agent configures agents, models, and tools with YAML or environment variables; mini-swe-agent emphasizes a small scaffold and linear message history.
- trae-agent supports several hosted providers and Ollama; mini-swe-agent lists integrations including LiteLLM, OpenRouter, and Portkey.
- Both can run tasks in Docker; mini-swe-agent also lists Podman, Singularity/Apptainer, bubblewrap, and contree environments.
- trae-agent records trajectories for debugging and analysis; mini-swe-agent includes a trajectory browser and batch inference for SWE-bench.
- Both are MIT-licensed Python projects. trae-agent requires Python 3.12 or newer for its documented setup and has measured quiet activity; mini-swe-agent identifies its repository as v2 and provides a migration guide for v1 users.
Choose trae-agent if you…
- need configurable tools and agent settings for multi-step software tasks.
- want to inspect recorded agent steps and model or tool interactions.
- want to run with a supported provider such as Ollama or use Docker execution.
Choose mini-swe-agent if you…
- want a compact agent scaffold that acts through bash.
- need Python bindings or a linear message history that is easy to inspect.
- want to run tasks across the listed environments or use batch inference for SWE-bench.
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.