Agents · Princeton NLP

SWE-agent

Research agent that pioneered agent-computer interfaces

[ Overview ]

SWE-agent (from Princeton NLP) is an open-source framework that lets language models autonomously fix GitHub issues using a standardized agent-computer interface for shells, editors, and test runners.

[ Pros ]

Influential agent-computer interface design
Lightweight, hackable codebase
Strong SWE-bench results

[ Cons ]

Requires technical setup
No managed hosting or UI

[ Tags ]

open-sourceresearchswe-bench