OpenHands
Open-source coding agent: self-host free or run in OpenHands Cloud
Overview
OpenHands (formerly OpenDevin) is an MIT-licensed autonomous coding agent from All Hands AI that plans, edits, runs, and tests code in a sandbox. It scores 53-77% on SWE-bench Verified depending on the model, with free self-hosting or a paid OpenHands Cloud/Enterprise tier.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 11 Jul 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- Web GUI, terminal UI, CLI
- Git integrations
- Model-agnostic
- Community support
- Up to 10 conversations/day
- BYOK or at-cost model pricing
- Jira and Slack integrations
- API for automation
- SAML/SSO
- Unlimited concurrent conversations
- Multi-user RBAC
- Priority support
Use cases
What you can produce with OpenHands
- Sandboxed autonomous coding agent (plans, edits, runs, tests code)
- MIT-licensed, self-hostable core
- Web GUI, terminal UI, and CLI
- Model-agnostic — BYOK or OpenHands-hosted models
- GitHub/GitLab/Bitbucket plus Jira/Slack integrations
- OpenHands Cloud with free tier and at-cost model pricing
- Enterprise: SSO/SAML, private VPC, multi-user RBAC
ASEAN Perspective
OpenHands in Southeast Asia
ASEAN-region availability and pricing notes coming soon. Drop the editorial team a note via /contact/ if you can supply local context (Singapore/Malaysia/Indonesia/Thailand/Vietnam).
Reading a repo, writing and running code, executing shell commands, browsing the web to close out a dev task end to end — that's the sandboxed loop this open-source agent runs, and it does so in the open: MIT-licensed, model-agnostic, and inspectable rather than a black box rented by the token. On SWE-bench Verified it scores anywhere from 53% to 77% depending on which model sits behind it, competitive with the closed alternatives, and the team's January 2026 OpenHands Index broadens evaluation past bug-fixing into greenfield builds, frontend work, and testing.
It's the right call for teams that want control over models, data, and infrastructure and are willing to spend setup time to get it — self-hosting plus per-task model costs add up, and like every autonomous agent it needs supervision; it will occasionally charge confidently in the wrong direction before you can stop it. 70k+ GitHub stars, active development, solid API/SDK and docs, works anywhere including self-hosted in ASEAN.
What people say
Few open coding agents have built as large a following: 70,000+ GitHub stars and roughly 500 contributors as of mid-2026, with some trackers citing figures as high as 76-80k depending on when they last synced. The project rebranded from OpenDevin in late 2024 under the All Hands AI banner and has since raised an $18.8M Series A.
The benchmark story is the strongest selling point. On SWE-bench Verified — the standard test for whether an agent can actually resolve real GitHub issues — OpenHands paired with a strong model like Claude 4.5 resolves over half of tasks, with reported scores ranging 53-77% depending on the underlying model. In January 2026 the team launched the OpenHands Index, extending evaluation beyond bug-fixing to greenfield app development, frontend tasks, and testing — an acknowledgment that SWE-bench alone doesn't capture everything developers use these agents for.
Third-party review aggregators put user ratings around 4.6/5, generally crediting the sandboxed execution model and breadth of integrations (Git, Jira, Slack). The most consistent criticism, echoed across review sites, is a variant of "free and open doesn't mean easy": self-hosting has real setup overhead, and the standard failure mode for any autonomous agent applies here too — it can head off in the wrong direction and make a mess of a codebase before a human notices and intervenes.
Cost comparisons favor OpenHands sharply when paired with open models: running OpenHands or a similar open agent on top of Qwen3-Coder or DeepSeek V3.2 has been benchmarked at around 5% of what an equivalent task costs on a closed agent like Devin, which is the core pitch for teams price-sensitive at scale. The tradeoff is that open-model quality still trails frontier closed models on harder tasks, so the savings come with a capability gap that narrows but hasn't closed.
For teams wanting a coding agent they can run, inspect, and swap models on rather than subscribe to, OpenHands is the most credible mainstream option in that lane.
Summary of public user & expert reviews, compiled by RECATOOLS.
About this listing
This entry was compiled from publicly available data including OpenHands's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with OpenHands unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to OpenHands directly →
Spotted something out of date? Suggest an update →
OpenHands in the news
More in Code & Dev Tools
