← Back to AI Agents
SWE-Agent by Princeton NLP is an open-source autonomous agent designed to resolve real-world GitHub Issues. It achieved top results on SWE-Bench and provides a foundation for building custom coding agents.
55%
任务成功率
240s
平均响应时间
200,000
Tokens/Task
详细评分
自主性
78/100
可靠性
58/100
易用性
45/100
成本效益
95/100
集成能力
50/100
支持
40/100
能力
CodeShellFile
主要特点
优点
- Completely free and open-source
- Top SWE-Bench performance
- Customizable agent framework
- Academic research quality
缺点
- Requires technical setup
- 55% success rate on real tasks
- No hosted version
- Requires own API keys
定价方案
Open Source
Free
- Full agent code
- Self-hosted
- Bring your own API keys
适用场景
Open-source DevelopmentResearchIssue ResolutionCustom Agent Building
成本模拟器
根据使用量估算您的月度成本
105001000
基础订阅$0
Token费用（100个任务）$88.00
月度总计$88.00
Underlying Models
GPT-4.1$2/$8 per 1M tokens
~200,000 tokens/task avg
查看所有代理定价
→Underlying Models
GPT-4.1
OpenAI · $2/$8 per 1M tokens
Claude 4 Sonnet
Anthropic · $3/$15 per 1M tokens
View Model Pricing →测试日期2025-03
最后更新2025-06
自主级别Fully Autonomous
