The open platform for cloud coding agents - autonomous software development that stays transparent, model-agnostic, and under your control.
OpenHands is an open-source platform for building and running AI coding agents - software that can pick up an engineering task and actually finish it. It writes code, runs commands, browses documentation, and edits files across a whole repository, all inside a sandboxed environment a team controls. Where many rivals ship a closed product, OpenHands ships the agent itself, source and all.
The project began in early 2024 as OpenDevin, an open response to the wave of autonomous-coding-agent demos. It started, in the founders' telling, as a single text file on GitHub. Within roughly eighteen months it had gathered more than 65,000 stars, over 3 million downloads, and a company - All Hands AI - built to carry the work forward. The project was renamed OpenHands along the way, and its logo, two raised hands, still winks at both the name and the community behind it.
What OpenHands is really selling is not novelty but trust. Its agents run on whichever model a team prefers - Claude, GPT, or something fine-tuned in-house - and can be self-hosted, audited, and governed. In a market racing toward proprietary black boxes, that openness is the whole pitch.
“Powerful coding agents will remain open, transparent, and under your control.”
CEO Robert Brennan points to the unglamorous half of a developer's week - writing unit tests, bumping dependencies, chasing down documentation drift, resolving low-severity vulnerabilities. These "toil-oriented" tasks eat time and, conveniently, are exactly the kind of routine work AI handles well.
OpenHands targets that work first. In enterprise use, teams report code-maintenance backlogs cut by roughly half, vulnerability fixes dropping from days to minutes, and refactors running in parallel across hundreds of repositories.
The open project draws hundreds of individual developers and 500+ contributors. On the enterprise side, engineers at some of the largest names in tech run OpenHands in their own environments.
Companies whose engineers are cited as OpenHands users. Adoption figures per company announcements.
Autonomous coding is a crowded field - Cognition's Devin, GitHub Copilot, Cursor, Google's Jules, and research agents like SWE-agent and Aider all compete for the same work. OpenHands stakes out the open, self-hostable corner: read the code, run it anywhere, swap the model.
Bars scaled for visual comparison, not to a common axis. Figures approximate, per company and repository data.
OpenHands meets developers wherever they work - on a laptop, in a terminal, or across an entire organization's cloud.
The open-source autonomous agent that solves engineering tasks end-to-end inside a sandboxed environment. Formerly OpenDevin.
Hosted, model-agnostic platform to run agents at scale, with GitHub, GitLab, Bitbucket, Slack, and Jira integrations.
Command-line and desktop interfaces for running agents on your own machine - privacy and cost first.
A composable, extensible Python SDK for building and shipping production coding agents.
Source-available deployment with Docker sandbox isolation, role-based access controls, and audit trails.
Integration with AMD's Lemonade Server to run private, on-device coding agents on Ryzen AI PCs.
The OpenHands agent is free and open source. All Hands AI earns revenue where scale and governance matter: OpenHands Cloud (usage-based hosting) and a source-available enterprise edition that adds the security, access controls, and audit trails large teams require.
It is the classic open-core wager - let the community build trust and adoption, then sell the layer that turns one local agent into a system running across an entire organization.
Lead investor Madrona framed OpenHands as building "the operating system for agentic software development" - the interface, automations, and control layer between raw models and real engineering work.
That positioning matters as models commoditize: the durable value moves to the platform that runs them safely, at scale, on a team's own terms.
A professor, a startup engineer, and a PhD student joined forces in the early days of the project and turned it into All Hands AI.
Former senior engineer at Google and product/engineering leader at open-source companies including Fairwinds.
Carnegie Mellon professor and longtime open-source advocate in natural language processing and agents.
Began the project as a PhD student at UIUC building strong open-source agents; now leads AI research.
Lead: Madrona
With Menlo Ventures, Pillar VC, Obvious Ventures, Fujitsu Ventures, and Alumni Ventures.
Lead: Menlo Ventures
With Pillar VC, Betaworks, and Rebellion - plus angels including Hugging Face's Thomas Wolf, Cloudera's Jeff Hammerbacher, and PyTorch creator Soumith Chintala.
Total funding to date: approximately $23.8M.
OpenHands is an open-source, model-agnostic platform for building and running AI coding agents that complete entire software engineering tasks - writing code, running commands, and using tools inside a sandboxed environment. It is developed by the company All Hands AI.
Yes. The core agent is free and open source with 65,000+ GitHub stars. All Hands AI monetizes through OpenHands Cloud and a source-available enterprise edition.
Yes. The project launched as OpenDevin in 2024 and was later renamed OpenHands when the founders formed All Hands AI.
OpenHands is model-agnostic. It works with Claude, GPT, and custom fine-tuned models, so teams aren't locked into a single provider.
Individual developers and open-source contributors, plus engineers at large companies including AMD, Apple, Google, Amazon, NVIDIA, Netflix, TikTok, Mastercard, and VMware.