OpenHands is an open source AI coding agent you can self-host — a Devin alternative that reads your repository, writes and edits code across files, runs commands and tests, and opens pull requests with the LLM you choose.

What is OpenHands?

OpenHands is an open source AI coding agent, built by All Hands AI, that acts like a software engineer: given a task or GitHub issue, it reads your repository, edits code across multiple files, runs terminal commands, browses the web for docs, and opens a pull request. You run it locally, in your own cloud, or through OpenHands Cloud, with the LLM you choose.

What is OpenHands best for?

OpenHands suits developers and teams who want an autonomous coding agent they control — running it on their own machine or infrastructure with their own API keys, rather than sending code to a closed SaaS. It works best on well-specified, test-driven tasks: bug fixes, scoped features, refactors, and chores where success is verifiable.

What can OpenHands do?

Read a repo and make changes across many files, then run tests and open a pull request
Run terminal commands and execute code in a sandboxed environment
Browse the web to read documentation and APIs while it works
Connect to GitHub, GitLab, and Bitbucket to act on issues and push branches
Work with any major LLM — Claude, GPT, open models, or bring-your-own-key
Run as a CLI, a local GUI with REST API, a Python agent SDK, or hosted cloud
Scale to many concurrent agents via an enterprise Kubernetes deployment

Where does OpenHands fall short?

It struggles with ambiguous or underspecified tasks and can ship incorrect or partial solutions — its output needs human code review before merging.
Results depend heavily on the LLM you connect and the API tokens you spend; complex multi-file work with several test iterations can cost several dollars per task.
The enterprise multi-user features (RBAC, Slack/Jira/Linear integrations, the control plane) live under a separate commercial license, not the MIT core.

Is OpenHands free?

Yes — the core OpenHands platform is free and MIT-licensed, so you can self-host it at no software cost and pay only for the LLM tokens it consumes (or run a local model). OpenHands Cloud has a free individual tier with bring-your-own-key or at-cost model pricing, capped at a limited number of conversations per day; unlimited concurrent usage and the enterprise self-hosted deployment are custom-priced.

What does OpenHands replace?

OpenHands is a self-hosted, open source alternative to Devin, Cognition’s proprietary AI software engineer, and competes with hosted agents like Google’s Jules. It does the same autonomous plan-write-test-and-PR job, but you run it yourself with the model and infrastructure you choose, instead of paying for a closed SaaS.

FAQ

Is OpenHands open source? Yes. The core platform and Docker images are MIT-licensed. Only the separate enterprise/ directory — multi-user, RBAC, and team integrations — uses a commercial license.

Can I self-host OpenHands for free? Yes. Self-hosting the MIT core is free; you only pay for the LLM API tokens the agent uses, or run a local open model to avoid that too.

Is OpenHands a good Devin alternative? For developers who want control and transparency, yes — it’s open source, self-hostable, and model-agnostic. Devin offers a more polished managed experience, while OpenHands trades that for no vendor lock-in and no per-seat pricing.

What do I need to run OpenHands? Docker on your machine (or a server) and an API key for an LLM such as Claude or GPT — or a local model. It runs the agent in a sandboxed container and pushes changes to your Git provider.