OpenHands: An Open Platform for AI Software Developers as Generalist Agents
2024/07/23 by Xingyao Wang, Wang, Xingyao, Boxuan Li +47 · 3 voices · 236 citations
Computer Science · #Multi-Agent Systems and Negotiation #cs.AI #cs.CL #cs.SE
paper · pdf · doi:10.48550/arxiv.2407.16741
openalex publication_date 2024/07/23 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Abstract
Software is one of the most powerful tools that we humans have at our disposal; it allows a skilled programmer to interact with the world in complex and profound ways. At the same time, thanks to improvements in large language models (LLMs), there has also been a rapid development in AI agents that interact with and affect change in their surrounding environments. In this paper, we introduce OpenHands (f.k.a. OpenDevin), a platform for the development of powerful and flexible AI agents that interact with the world in similar ways to those of a human developer: by writing code, interacting with a command line, and browsing the web. We describe how the platform allows for the implementation of new agents, safe interaction with sandboxed environments for code execution, coordination between multiple agents, and incorporation of evaluation benchmarks. Based on our currently incorporated benchmarks, we perform an evaluation of agents over 15 challenging tasks, including software engineering (e.g., SWE-BENCH) and web browsing (e.g., WEBARENA), among others. Released under the permissive MIT license, OpenHands is a community project spanning academia and industry with more than 2.1K contributions from over 188 contributors.
Cited by
- AuditRepairBench: A Paired-Execution Trace Corpus for Evaluator-Channel Ranking Instability in Agent Repair
- Beyond Text-to-SQL: Can LLMs Really Debug Enterprise ETL SQL?
- Replication Package (v2): Output Format × Model Identity Interaction Effects
- FedAgentKE: Federated Semantic Knowledge Evolution for Heterogeneous Agents
- PerfAgent: Profiler-Guided Iterative Refinement for Repository-Level Code Optimization
- Know Your Agent: Reconnaissance-Driven Pentesting of AI Agents
- Beyond Fail-to-Pass: Iterative Hardening of Co-Generated Bug Reproduction Tests and Fixes
- PhoenixRepair: Rethinking Repair Strategy Exploration in Software Agents
- NVIDIA-labs OO Agents: Native Python Object-Oriented Agents
- IssueTrojanBench: Benchmarking AI Coding Agents Against Malicious Issue Requests
- FlashRT: Agent Harness for Guiding Agents to Deploy Real-Time Multimodal Applications
- AgentCompass: A Unified Evaluation Infrastructure for Agent Capabilities
- Knowledge-Centric Self-Improvement
- SWE-Milestone: Evaluating AI Agents on Continuous Software Evolution
- LLM-Based Invariant Testing for Software Functional Bugs
- Athena-Brain Technical Report: An Efficient Robot Brain for General Intelligence and Embodied Interaction
- CodeRescue: Budget-Calibrated Recovery Routing for Coding Agents
- RoboHarness: Memory-Driven Orchestration of Heterogeneous Robot Policies for Long-Horizon Planning
- How Agent Skills Fail under Long Contexts: A White-Box Study in Code Auditing
- DepRepair: LLM-Based Source-Code Repair for Dependency Breaking Changes
- Don't Blame the Large Language Model: How Agent Harness Evolution Shapes Coding Agent Quality
- TRIM: Reducing AI-Generated CodeSlop via Agent Trajectory Minimization
- SWE-Pruner Pro: The Coder LLM Already Knows What to Prune
- Agentic Vulnerability Reasoning on COTS Binaries
- Beyond Resolved Rate: A Non-Functional Quality Study
- Function-Aware Fill-in-the-Middle as Mid-Training for Coding Agent Foundation Models
- IssueExec: A Test-Driven Approach for Localizing Software Engineering Issues
- Position: Coding Benchmarks Are Misaligned with Agentic Software Engineering
- Evaluating LLM-Based 0-to-1 Software Generation in End-to-End CLI Tool Scenarios
- Workflow-GYM: Towards Long-Horizon Evaluation of Computer-use Agentic tasks in Real-World Professional Fields
- Pomona: Continuous Code Quality Improvement via Small, Agentic Pull Requests at Bloomberg
- Nonuniformity Principle in Human-AI Coworking
- Large-Scale Terminal Agentic Trajectory Generation from Dockerized Environments
- NEMO: Execution-Aware Optimization Modeling via Autonomous Coding Agents
- Are Performance-Optimization Benchmarks Reliably Measuring Coding Agents?
- Latent Programming Horizons in Coding Agents
- Self-Harness: Harnesses That Improve Themselves
- Continual Harness: Online Adaptation for Self-Improving Foundation Agents
- Agentic Harness Engineering: Observability-Driven Automatic Evolution of Coding-Agent Harnesses
- Constraint Decay: The Fragility of LLM Agents in Backend Code Generation
- How Do AI Agents Spend Your Money? Analyzing and Predicting Token Consumption in Agentic Coding Tasks
- Dive into Claude Code: The Design Space of Today's and Future AI Agent Systems
- AutoSurrogate: An LLM-driven multi-agent framework for autonomous construction of deep learning surrogate models in subsurface flow
- Evaluating AGENTS.md: Are Repository-Level Context Files Helpful for Coding Agents?
- Let It Flow: Agentic Crafting on Rock and Roll, Building the ROME Model within an Open Agentic Learning Ecosystem
- Huxley-Gödel Machine: Human-Level Coding Agent Development by an Approximation of the Optimal Self-Improving Machine
- AISSISTANT: Human-AI Collaborative Review and Perspective Research Workflows in Data Science
- GLM-4.5: Agentic, Reasoning, and Coding (ARC) Foundation Models
- Future of Work with AI Agents: Auditing Automation and Augmentation Potential across the U.S. Workforce
- Gradient-Based Program Repair: Fixing Bugs in Continuous Program Spaces
- Transforming Science with Large Language Models: A Survey on AI-assisted Scientific Discovery, Experimentation, Content Generation, and Evaluation
- Agentic Code Reasoning
- NERFIFY: A Multi-Agent Framework for Turning NeRF Papers into Code
- GraphLocator: Graph-guided Causal Reasoning for Issue Localization
- Agentic Software Issue Resolution with Large Language Models: A Survey
- From Cognitive Architectures to Language Agents: A Mechanism-Level Review of Lineage, Convergence, and Migration Gaps
- HANDBOOK.md: A Benchmark for Long-Context Agentic Instruction Following
- Matryoshka Agent: Unfolding Sub-Agents for Long-Horizon Machine Learning Engineering
- How Do Practitioners Build SE Agents? Insights from a Mixed-Methods Study
- Agent Team Work Zone: An Automated, Persistent Workspace for Long-Lived Claude Code Agent Teams
- Context as a Tool: Context Management for Long-Horizon SWE-Agents
- JETO-Bench: A Reproducible Benchmark for Execution Time Improvement Patches in Java
- TrajAudit: Automated Failure Diagnosis for Agentic Coding Systems
- SF-AMS: Strategic Forgetting for Structured Memory in LLM Agent
- SWE-RM: Execution-free Feedback For Software Engineering Agents
- C2LLM Technical Report: A New Frontier in Code Retrieval via Adaptive Cross-Attention Pooling
- One Tool Is Enough: Reinforcement Learning for Repository-Level LLM Agents
- ABBEL: Learning Natural-Language Belief States for Memory-Efficient Interaction
- MemEvolve: Meta-Evolution of Agent Memory Systems
- SWE-EVO: Benchmarking Coding Agents in Long-Horizon Software Evolution Scenarios
- Imitation Learning for Multi-turn LM Agents via On-policy Expert Corrections
- NL2Repo-Bench: Towards Long-Horizon Repository Generation Evaluation of Coding Agents
- Training Versatile Coding Agents in Synthetic Environments
- Confucius Code Agent: Scalable Agent Scaffolding for Real-World Codebases
- SWEnergy: An Empirical Study on Energy Efficiency in Agentic Issue Resolution Frameworks with SLMs
- Autonomous Issue Resolver: Towards Zero-Touch Code Maintenance
- Reformulate, Retrieve, Localize: Agents for Repository-Level Bug Localization
- Nex-N1: Agentic Models Trained via a Unified Ecosystem for Large-Scale Environment Construction
- DAComp: Benchmarking Data Agents across the Full Data Intelligence Lifecycle
- InnoGym: Benchmarking the Innovation Potential of AI Agents
- Process-Centric Analysis of Agentic Software Systems
- Is Vibe Coding Safe? Benchmarking Vulnerability of Agent-Generated Code in Real-World Tasks
- Simple Agents Outperform Experts in Biomedical Imaging Workflow Optimization
- LeechHijack: Covert Computational Resource Exploitation in Intelligent Agent Systems
- BackportBench: A Multilingual Benchmark for Automated Backporting of Patches
- SABER: Small Actions, Big Errors -- Safeguarding Mutating Steps in LLM Agents
- LLM Assisted Coding with Metamorphic Specification Mutation Agent
- SkyRL-Agent: Efficient RL Training for Multi-turn LLM Agent
- InfCode-C++: Intent-Guided Semantic Retrieval and AST-Structured Search for C++ Issue Resolution
- InfCode: Adversarial Iterative Refinement of Tests and Patches for Reliable Software Issue Resolution
- Agent0: Unleashing Self-Evolving Agents from Zero Data via Tool-Integrated Reasoning
- Rethinking Kernel Program Repair: Benchmarking and Enhancing LLMs with RGym
- Live-SWE-agent: Can Software Engineering Agents Self-Evolve on the Fly?
- Beyond Accuracy: Behavioral Dynamics of Agentic Multi-Hunk Repair
- AlphaResearch: Accelerating New Algorithm Discovery with Language Models
- MURPHY: Multi-Turn GRPO for Self Correcting Code Generation
- Self-Abstraction from Grounded Experience for Plan-Guided Policy Refinement
- Klear-AgentForge: Forging Agentic Intelligence through Posttraining Scaling
- SWE-fficiency: Can Language Models Optimize Real-World Repositories on Real Workloads?
- SWE-Compass: Towards Unified Evaluation of Agentic Coding Abilities for Large Language Models
- ArchPilot: A Proxy-Guided Multi-Agent Approach for Machine Learning Engineering
- Collaborative Agents for Automated Program Repair in Ruby
- The OpenHands Software Agent SDK: A Composable and Extensible Foundation for Production Agents
- SWE-Sharp-Bench: A Reproducible Benchmark for C# Software Engineering Tasks
- Training Proactive and Personalized LLM Agents
- AgentGit: A Version Control Framework for Reliable and Scalable LLM-Powered Multi-Agent Systems
- Interaction as Intelligence Part II: Asynchronous Human-Agent Rollout for Long-Horizon Task Training
- Understanding Code Agent Behaviour: An Empirical Study of Success and Failure Trajectories
- Gistify! Codebase-Level Understanding via Runtime Execution
- Empowering RepoQA-Agent based on Reinforcement Learning Driven by Monte-carlo Tree Search
- The FM Agent
- QCoder Benchmark: Bridging Language Generation and Quantum Hardware through Simulator-Based Feedback
- Completion ≠ Collaboration: Scaling Collaborative Effort with Agents
- Process-Level Trajectory Evaluation for Environment Configuration in Software Engineering Agents
- How Do LLMs Read Bug Reports? An Empirical Study of Attention in LLMs for Automated Program Repair
- OrchBench: Evaluating Multi-Agent Orchestration Plans in Isolation via Deterministic Simulation
- Messier: A High-Resolution Corpus for Cross-Benchmark Agent Evaluation
- BM25 Wins at Scale: A Scaling Study of Retrieval-Augmented Generation Paradigms
- ExplainBench: Evaluating Code Explanations from Agents
- DREvo: Distilling Recalibrated Historical Experience for Harness Self-Evolution
- Rethinking Self-Evolution: A Constrained Exploration-Exploitation Process for Mitigating Skill Overfitting
- CodeSpec: Dual Executable Specifications for Agentic Long-Horizon Feature Development
- ALMAS: an Autonomous LLM-based Multi-Agent Software Engineering Framework
- Agents' Last Exam
- Nautilus: From One Prompt to Plug-and-Play Robot Learning
- CooperBench: Why Coding Agents Cannot be Your Teammates Yet
- Sharpness-Guided Group Relative Policy Optimization via Probability Shaping
- Agent Data Protocol: Unifying Datasets for Diverse, Effective Fine-tuning of LLM Agents
- Towards AI as Colleagues: Multi-Agent System Improves Structured Professional Ideation
- TDFlow: Agentic Workflows for Test Driven Software Engineering
- Alita-G: Self-Evolving Generative Agent for Agent Generation
- Scalable Supervising Software Agents with Patch Reasoner
- How Do AI Agents Do Human Work? Comparing AI and Human Workflows Across Diverse Occupations
- TOM-SWE: User Mental Modeling For Software Engineering Agents
- Automated Cloud Infrastructure-as-Code Reconciliation with AI Agents
- Code-enabled language models can outperform reasoning models on diverse tasks
- BugPilot: Complex Bug Generation for Efficient Learning of SWE Skills
- TheMCPCompany: Creating General-purpose Agents with Task-specific Tools
- KAT-Coder Technical Report
- EvoSyn: Generalizable Evolutionary Data Synthesis for Verifiable Learning
- SIADAFIX: issue description response for adaptive program repair
- LLM Agents for Automated Web Vulnerability Reproduction: Are We There Yet?
- E2Edev: Benchmarking Large Language Models in End-to-End Software Development Task
- When "Correct" Is Not Safe: Can We Trust Functionally Correct Patches Generated by Code Agents?
- Benchmarking Correctness and Security in Multi-Turn Code Generation
- Lingxi: Repository-Level Issue Resolution Framework Enhanced by Procedural Knowledge Guided Scaling
- Scaling Long-Horizon LLM Agent via Context-Folding
- BrowserAgent: Building Web Agents with Human-Inspired Web Browsing Actions
- On the Scaling of PEFT: Towards Million Personal Models of Trillion Parameters
- MUSE-Autoskill: Self-Evolving Agents via Skill Creation, Memory, Management, and Evaluation
- Effective Strategies for Asynchronous Software Engineering Agents
- How can we assess human-agent interactions? Case studies in software agent design
- FlowSearch: Advancing deep research with dynamic structured knowledge flow
- Past, Present, and Future of Bug Tracking in the Generative AI Era
- Test-Time Reasoners Are Strategic Multiple-Choice Test-Takers
- Learning on the Job: An Experience-Driven Self-Evolving Agent for Long-Horizon Tasks
- AutoMLGen: Navigating Fine-Grained Optimization for Coding Agents
- CUDA Agent: Large-Scale Agentic RL for High-Performance CUDA Kernel Generation
- FreshBrew: A Benchmark for Evaluating AI Agents on Java Code Migration
- SpikingMamba: Towards Energy-Efficient Large Language Models via Knowledge Distillation from Mamba
- Retrieval-Augmented Code Generation: A Survey with Focus on Repository-Level Approaches
- BrowserArena: Evaluating LLM Agents on Real-World Web Navigation Tasks
- REFINE: Enhancing Program Repair Agents through Context-Aware Patch Refinement
- LLM Agents for Automated Dependency Upgrades
- Breaking the Code: Security Assessment of AI Code Agents Through Systematic Jailbreaking Attacks
- Improving Code Localization with Repository Memory
- Lita: Light Agent Uncovers the Agentic Coding Capabilities of LLMs
- BloomAPR: A Bloom's Taxonomy-based Framework for Assessing the Capabilities of LLM-Powered APR Solutions
- Towards Reliable Generation of Executable Workflows by Foundation Models
- Agentic Services Computing
- TENET: Leveraging Tests Beyond Validation for Code Generation
- PerfBench: Can Agents Resolve Real-World Performance Bugs?
- MCPMark: A Benchmark for Stress-Testing Realistic and Comprehensive MCP Use
- HFuzzer: Testing Large Language Models for Package Hallucinations via Phrase-based Fuzzing
- Improving the Efficiency of LLM Agent Systems through Trajectory Reduction
- WebGen-Agent: Enhancing Interactive Website Generation with Multi-Level Feedback and Step-Level Reinforcement Learning
- EPO: Entropy-regularized Policy Optimization for LLM Agents Reinforcement Learning
- SecureAgentBench: Benchmarking Secure Code Generation under Realistic Vulnerability Scenarios
- A Benchmark for Localizing Code and Non-Code Issues in Software Projects
- Efficient On-Device Agents via Adaptive Context Management
- An Empirical Study of Coordination Mode as the First-Class Citizen in From-Scratch Multi-Agent Coding
- FAVA: Formal Authorization for Verified Agents with Evidence-Backed Permission Graphs
- RLPF: Reinforcement Learning from Performance Feedback for Code Generation
- SWE-NFI: Studying and Benchmarking Coding Agents for Non-Functional Improvements
- OwlPath: Lossless Knowledge Compression for LLM Bug Repair
- Evaluating LLM Agents on Automated Software Analysis Tasks
- Generalizable End-to-End Tool-Use RL with Synthetic CodeGym
- SWE-Bench Pro: Can AI Agents Solve Long-Horizon Software Engineering Tasks?
- RPG: A Repository Planning Graph for Unified and Scalable Codebase Generation
- MatchFixAgent: Language-Agnostic Autonomous Repository-Level Code Translation Validation and Repair
- OpenLens AI: Fully Autonomous Research Agent for Health Infomatics
- SWE-QA: Can Language Models Answer Repository-level Code Questions?
- Who is Introducing the Failure? Automatically Attributing Failures of Multi-Agent Systems via Spectrum Analysis
- An Empirical Study on Failures in Automated Issue Solving
- Agentic AI for Software: thoughts from Software Engineering community
- Building Coding Agents via Entropy-Enhanced Multi-Turn Preference Optimization
- CodeCureAgent: Automatic Classification and Repair of Static Analysis Warnings
- SWE-Effi: Re-Evaluating Software AI Agent System Effectiveness Under Resource Constraints
- EnvX: Agentize Everything with Agentic AI
- Agentic Software Engineering: Foundational Pillars and a Research Roadmap
- AgenTracer: Who Is Inducing Failure in the LLM Agentic Systems?
- app.build: A Production Framework for Scaling Agentic Prompt-to-App Generation with Environment Scaffolding
- LLM Agents for Generating Microservice-based Applications: how complex is your specification?
- VerlTool: Towards Holistic Agentic Reinforcement Learning with Tool Use
- Reinforcement Learning for Machine Learning Engineering Agents
- Reflective Paper-to-Code Reproduction Enabled by Fine-Grained Verification
- The Complexity Trap: Simple Observation Masking Is as Efficient as LLM Summarization for Agent Context Management
- ASIC-Agent: An Autonomous Multi-Agent System for ASIC Design with Benchmark Evaluation
- AI and Agile Software Development: A Research Roadmap from the XP2025 Workshop
- Reliable Weak-to-Strong Monitoring of LLM Agents
- GitTaskBench: A Benchmark for Code Agents Solving Real-World Tasks Through Code Repository Leveraging
- MedResearcher-R1: Expert-Level Medical Deep Researcher via A Knowledge-Informed Trajectory Synthesis Framework
- ComputerRL: Scaling End-to-End Online Reinforcement Learning for Computer Use Agents
- Illuminating LLM Coding Agents: Visual Analytics for Deeper Understanding and Enhancement
- LinkAnchor: An Autonomous LLM-Based Agent for Issue-to-Commit Link Recovery
- You Don't Know Until You Click:Automated GUI Testing for Production-Ready Software Evaluation
- AI Agentic Programming: A Survey of Techniques, Challenges, and Opportunities
- Position: Intelligent Coding Systems Should Write Programs with Justifications
- AutoIAD: Manager-Driven Multi-Agent Collaboration for Automated Industrial Anomaly Detection
- Empowering Time Series Forecasting with LLM-Agents
- Training Long-Context, Multi-Turn Software Engineering Agents with Reinforcement Learning
- Tool-integrated Reinforcement Learning for Repo Deep Search
- Polymath: A Self-Optimizing Agent with Dynamic Hierarchical Workflow
- A Multi-Agent System for Complex Reasoning in Radiology Visual Question Answering
- A Survey on AgentOps: Categorization, Challenges, and Future Directions
- RepoForge: Training a SOTA Fast-thinking SWE Agent with an End-to-End Data Curation Pipeline Synergizing SFT and RL at Scale
- Leveraging Large Language Model for Information Retrieval-based Bug Localization
- Multi-Agent Game Generation and Evaluation via Audio-Visual Recordings
- Measuring Harmfulness of Computer-Using Agents
- Trae Agent: An LLM-based Agent for Software Engineering with Test-time Scaling
- SWE-Exp: Experience-Driven Software Issue Resolution
- SWE-Debate: Competitive Multi-Agent Debate for Software Issue Resolution
- AutoCodeSherpa: Symbolic Explanations in AI Coding Agents
- Can Language Models Discover Scaling Laws?
- Prometheus: Unified Knowledge Graphs for Issue Resolution in Multilingual Codebases
- Agentic Program Repair from Test Failures at Scale: A Neuro-symbolic approach with static analysis and test execution feedback
Discussions
Related