Rating certificate · AW/26-09/66126 Sept 2026
Not enough evidence
The Register / Coding agents / OpenHands
Register entry · Coding agents · updated 26 Sept 2026
OpenHands
OpenHands (originally OpenDevin) is an MIT-licensed platform for autonomous software agents that read code, edit files, run commands and open pull requests. It ships a local web GUI that runs the agent in a Docker sandbox, a terminal CLI, a Python Software Agent SDK, and a hosted OpenHands Cloud with Jira/Slack integrations. An Enterprise tier offers SaaS or self-hosted-in-VPC deployment; the company raised an $18.8M Series A in November 2025.
What reviewers and users praise
- MIT-licensed and model-agnostic, including local models
- Sandboxed execution by default in the GUI
- Strong historical SWE-bench results and an active release cadence
What they criticise
- Docker-based setup is heavier than single-binary CLIs
- Users report UI performance and stability bugs in the Agent Canvas
- No official IDE extension
- Cloud privacy policy allows training on user content
Experts vs the crowdneeds both scores
Experts—
Crowd—
| Price | Free |
|---|---|
| Use it via | Web app, CLI, Desktop app, API, Chat apps |
| Runs on | macos, linux, windows, web |
| Licence | MIT |
| Updated | Evidence collected 26 Sept 2026 |
How the rating was worked out
Weights for Coding agents: adoption 35%, experts 40%, crowd 25%. Fewer than two pillars have enough evidence. The method.
| Adoption & momentumweight 35% | 76 | GitHub stars 89k → 89 · PyPI monthly downloads 173k → 43 · 38 releases in 90 days → 100reach × 0.7 + momentum × 0.3 |
|---|---|---|
| Expertsindependent reviews · weight 40% | — | Only 2 independent reviews (3 needed) Benchmarks, shown not scored: Terminal-Bench 2.0 No. 35 of 43 (OpenHands + Claude Opus 4.5) · SWE-bench Verified No. 16 of 77 (OpenHands + GPT-5) · SWE-bench Lite No. 19 of 58 (OpenHands + CodeAct v2.1 (claude-3-5-sonnet-20241022)) · SWE-bench Test No. 6 of 13 (OpenHands + CodeAct v2.1 (claude-3-5-sonnet-20241022)) · SWE-bench Multimodal No. 4 of 12 (OpenHands-Versa + Claude Sonnet 4) |
| Crowdcommunity sample · weight 25% | — | Only 5 community items sampled (6 needed) |
| Rating | — | Not rated: fewer than two pillars have enough evidence. |
Evidence checks. 3 GitHub issues not counted as sentiment. How evidence is checked.
Trust & safety
Seven dimensions graded A–D from documented facts. Overall: B.
A
Permission model
Local GUI runs the agent in a Docker container by default (RUNTIME=docker (default)); the non-isolated process runtime is described as unsafe. CLI asks for confirmation by default with opt-in --always-approve and --llm-approve modes; SDK documents AlwaysConfirm/NeverConfirm/ConfirmRisky policies. source
B
Data access scope
Agent operates inside the sandbox on the mounted workspace/repository; cloud connects via Git provider integrations. Broad within the workspace; scoping controls not separately documented in what was read. source
A
Data storage
Open-source edition runs locally; Enterprise can be self-hosted in the customer's VPC. Cloud is optional. source
D
Data retention & training
Privacy policy (effective 2025-09-03) for the hosted Services says 'we may use content and feedback from your use of the Services to train and tune our AI models'; retention is 'as long as reasonably necessary'. Applies to OpenHands Cloud, not self-run OSS. source
C
Incident history (24 months)
Two command-injection CVEs in 2026, both addressed in later versions. source
B
Compliance
Vanta-hosted trust center exists; a third-party review reports the vendor's July 2026 Security FAQ lists SOC 2 Type II as in progress and ISO 27001 not held. Trust center page did not render for direct verification. source
A
Transparency
MIT-licensed core with detailed public GitHub release notes. source
The experts
2 named independent reviews found; dated reviews from the last 12 months are scored.
- mixed
Not IDE-native by default — there is no official VS Code or JetBrains extension; only an unofficial, community-built VS Code integration exists
Hans Kuepper (PromptQuorum) · 5 Sept 2026 · promptquorum.com - positive
Enterprise-ready platform with web UI, multi-agent architecture, and production features
LocalAIMaster · 6 Feb 2026 · localaimaster.com
The crowd
5 coded posts from hackernews.
1 positive2 mixed2 negative
- positive
Customers deposit a project, sandboxed coding agents (OpenHands) build it enriched by my knowledge (10y of eng. practices), a demo goes live on a subdomain
Hacker News comment · 1 Sept 2026 · news.ycombinator.com - negative
There's OpenHands, which is more like Codex/Claude Code web UI. You run it locally and connect to remote LLMs. Kinda clunky, limited, poor design.
Hacker News comment · 30 Apr 2026 · news.ycombinator.com - negative
Got rid of openhands in favor of Shelley, stuff like that.
Hacker News comment · 27 Jan 2026 · news.ycombinator.com - mixed
your OpenHands concept with container per agent seem little heavy.
Hacker News comment · 16 Jun 2026 · news.ycombinator.com - mixed
OpenHands has self-hosted, KiloCode and RooCode both have a cloud option. I don't think you are able to pass a session around with any of them.
Hacker News comment · 7 Apr 2026 · news.ycombinator.com
Pricing
As published, checked 26 Sept 2026.
| Plan | Price | Notes |
|---|---|---|
| Open Source (Local) | Free | Single user, runs locally; GUI, TUI, CLI; bring your own model key. |
| Individual (Cloud) | Free | Hosted access, 1 user, 10 conversations/day max; BYOK or pay-as-you-go OpenHands LLM provider billed at cost. |
| Enterprise | Custom | SaaS or self-hosted in your VPC, SAML/SSO, unlimited users/concurrency, named customer engineer. |
Recent changes
- 25 Sept 2026
v1.24.0: workspace folder toggles in Conversations header; shared automation conversations open read-only on cloud. source
- 17 Sept 2026
v1.20.0: agent profiles can select available secrets; automations can use saved agent profiles. source
- 27 Aug 2026
v1.16.0: Linux desktop installer build; selectable LLM providers; live phase for automation runs. source
- 29 Jul 2026
v1.7.0: persistent agent memory toggle; LLM selector always shown. source
Ranked alongside
Other ranked entries in Coding agents.
OpenCodeAnomaly75
Claude CodeAnthropic73
Zed (Agent Panel)Zed Industries73
GitHub CopilotGitHub (Microsoft)72
Codex (CLI and app)OpenAI71
Sources
Every figure above traces to one of these, observed up to 26 Sept 2026.
- OpenHands SDK: Security & Action Confirmation
- Third-party page quoting the vendor's July 2026 Security FAQ (SOC 2 Type II in progress)
- HN post: OpenHands raised $18.8M Series A led by Madrona (Nov 2025)
- OpenHands Software Agent SDK (1,169 stars; 35 releases since 2026-06-28; PyPI openhands-sdk 4.5M downloads/month)