"Our key insight is that LMs require carefully designed agent-computer interfaces (similar to how humans like  good UI design). E.g. When the LM messes up indentation, our editor prevents it and gives logo

"Our key insight is that LMs require carefully designed agent-computer interfaces (similar to how humans like good UI design). E.g. When the LM messes up indentation, our editor prevents it and gives

Free

Open-source AI agent that autonomously fixes GitHub issues

FreeFree tier
Type
Open Source
Company
Princeton NLP

About "Our key insight is that LMs require carefully designed agent-computer interfaces (similar to how humans like good UI design). E.g. When the LM messes up indentation, our editor prevents it and gives

SWE-agent is an open-source system developed by Princeton NLP that autonomously resolves issues in GitHub repositories. It uses a carefully designed agent-computer interface—a specialized terminal—that allows GPT-4 to efficiently edit and run code. The system achieves accuracy comparable to Devin on the SWE-bench benchmark, averaging just 93 seconds per task. Key capabilities include opening, scrolling, and searching files; editing specific lines with automatic syntax checking; and writing and executing tests. The custom interface is critical for performance, preventing common LM errors (e.g., indentation mistakes). SWE-agent is easily configurable and extendable, making it a valuable tool for both software maintenance and research on software engineering agents.

Key Features

Autonomously solves issues in GitHub repositories
Custom agent-computer interface (specialized terminal) for GPT-4
Open, scroll, and search through files
Edit specific lines with automatic syntax checking
Write and execute tests
Similar accuracy to Devin on SWE-bench
Average resolution time of 93 seconds
Open source and easily configurable/extendable

Pros & Cons

Pros
  • Open source with active community contributions
  • High accuracy on SWE-bench, comparable to state-of-the-art systems
  • Fast average resolution time (93 seconds)
  • Custom interface reduces common LM errors like indentation mistakes
  • Easily configurable and extendable for further research
Cons
  • Requires GPT-4, a proprietary and costly model
  • Limited to code editing tasks within GitHub repositories
  • May not handle complex software engineering tasks beyond code changes

Best For

Fixing bugs and resolving issues in GitHub repositoriesAutomating software maintenance tasksResearch and development of software engineering agents

FAQ

What is SWE-agent?
SWE-agent is an open-source system for autonomously solving issues in GitHub repositories using a custom agent-computer interface.
How fast does SWE-agent resolve issues?
On SWE-bench, it averages 93 seconds per issue.
Is SWE-agent open source?
Yes, it is open source and available on GitHub. Anyone can experiment with and contribute new ways for agents to interact with computers.
What language model does SWE-agent use?
It is designed to work with GPT-4, leveraging a specialized terminal interface for optimal performance.
How does SWE-agent differ from simply connecting a language model to a terminal?
SWE-agent uses a carefully designed agent-computer interface (similar to good UI design for humans) that prevents common errors, such as indentation mistakes, and provides feedback.