Skip to content
Rate AI Agent
Open menu

SWE-agent

by SWE-agent team (Princeton & Stanford)

Research agent that turns GitHub issues into fixes

Not evaluatedNo independent result on a full benchmark suite yet.Website(opens in a new tab)Docs(opens in a new tab)Source(opens in a new tab)
Compare with…

Overview

Publisher-provided · source: github.com · compiled 15 Oct 2026

An academic open-source agent that takes a GitHub issue and attempts to resolve it using a purpose-built agent-computer interface.

Capabilities

  • code editing
  • multi file changes
  • terminal commands
  • codebase search
  • issue resolution

Benchmark results

Measured task performance on a published benchmark suite. Each metric is shown separately.

Not evaluatedNo independent result on a full benchmark suite yet.

This agent has not been evaluated yet.

We show no score rather than estimating one. Agents are added to the evaluation queue as benchmark capacity allows.

How evaluations work

Publisher claims

What the developer says about performance. Kept separate from measured results and not verified by us.

No performance claims recorded.

Community reports

Experiences reported by users. Not verified.

No community reports yet.

Known limitations

None documented yet. Limitations are added from evaluation failure analysis and verified reports.