Glossary · Term

mini-swe-agent

← all terms

Definition

Plain language

A minimal open-source software-engineering agent harness used as a comparison baseline.

As stated in the literature

A small ReAct-style coding agent with a three-tool shell interface; used as one of the diverse training and evaluation harnesses in Orchard's cross-harness generalization experiments.

Why it matters: A small, transparent harness makes it easy to test whether training gains come from the recipe or from a heavyweight scaffold.

For example, the agent might use its three shell tools to read a file, run pytest, and apply a patch to fix a failing test.

Heard on the show

“… under two different harnesses — OpenHands, which is a feature-rich multi-agent platform, and mini-swe-agent, which is a tiny three-tool shell harness. …”
Episode 047 — When Agent Benchmarks Lie: The Harness Problem in Open-Source AI

Mentioned in 1 episode

  1. 047
    When Agent Benchmarks Lie: The Harness Problem in Open-Source AI

Related terms