> ## Content Index > Fetch the complete content index at: https://www.theaidirectory.com.au/llms.txt > Use this file to discover other available public pages before exploring further. # OpenHands Review Australia 2026: The Open-Source Coding Agent, Judged - URL: https://www.theaidirectory.com.au/review-openhands-australia-2026/ - Published: 2026-06-27T23:00:00.000Z - Updated: 2026-10-09T06:34:03.000Z - Description: OpenHands is a powerful free coding agent — but Australian teams carry the data-residency work themselves. - Author: Adam - Tags: Coding & Development, #review **16**/25 ★★★★☆ Very Good OpenHands (formerly All Hands AI) Verified 29 Sep 2026 Reviewed 28 Jun 2026 Value for Money**5**/5 Ease of Use**2**/5 Feature Depth**5**/5 Australian Fit**2**/5 Reliability & Support**2**/5 **The gate:** Five stars need 21+ and Australian Fit of at least 4/5. [Skip to the verdict ](#verdict) *Pricing figures updated 6 September 2026: USD prices are converted at the current rate (1 USD ≈ 1.40 AUD, verified 5 September 2026) and shown ex-GST, replacing the retired 1.55 + GST basis. A GST-registered Australian business that supplies its ABN is not charged GST on an imported digital service; other buyers should allow \~10% on top (*[*Methodology*](https://www.theaidirectory.com.au/methodology/)*).* > **Verdict (TL;DR):** OpenHands is one of the most capable free coding agents you can run. It plans, writes, and applies changes across a whole codebase, not just autocomplete. You can self-host it or use the free hosted tier. For Australian developers who want data control, the self-host path is genuinely useful. The catch: setup is real work, the cloud tier caps you at 10 conversations a day, and there is no AUD billing, no AU support, and no published Australian hosting region. Best for technical teams who will configure their own infrastructure. > > **Score: 16/25 — ⭐⭐⭐⭐ Very Good** > > *Last reviewed: 28 June 2026 by Adam.* --- ## What OpenHands is OpenHands is an open-source autonomous coding agent. It does not just suggest code. It plans a task, writes the code, runs commands, and applies changes across your codebase end to end. The project is published by OpenHands, the company formerly trading as All Hands AI. The core agent is released under the MIT licence. As of late June 2026, the GitHub repository carries more than 78,000 stars, per the vendor's own GitHub page. The product runs in three shapes. You can run it locally on your own machine. You can use OpenHands Cloud, the hosted version. Or you can deploy OpenHands Enterprise self-hosted inside your own infrastructure. Like other open agents, OpenHands does not own the model layer. It is model-agnostic. You bring your own LLM key, or you buy inference through the OpenHands provider at cost. The agent works with Anthropic, OpenAI, Google, and others. What sets it apart is breadth. The vendor positions OpenHands as a control centre for coding agents. Its own documentation describes running the OpenHands agent or third-party agents such as Claude Code, Codex, and Gemini through one interface. It also supports parallel agent execution, scheduling, and webhook-triggered automations. In short, this is not a line-by-line assistant. It is a full agent platform aimed at real engineering work: bug fixes, pull-request reviews, code migration, and incident triage. --- ## What it costs in Australia There are three tiers on the OpenHands pricing page, captured 30 June 2026. **Open Source (local) — free.** This is the self-hosted agent. You get the web GUI, the terminal UI, the CLI, Git integrations, and community support. You run it on your own machine or infrastructure. There is no licence fee. **Individual (cloud) — free.** This is the hosted SaaS tier. It adds cloud access on desktop and mobile, API support, and Jira and Slack integrations. It is limited to a single user and a maximum of 10 conversations per day, per the vendor's pricing page. For LLM costs, the vendor states you "bring your own key or use our providers at-cost," billed pay-as-you-go with no markup. **Enterprise (SaaS or self-hosted) — custom pricing.** This adds SAML and SSO, unlimited concurrent conversations per user, a large-codebase SDK, and priority support via a shared Slack channel. Pricing is quote-only; the vendor lists no public figure. So the tool itself is free on two of three tiers. Your real spend is LLM inference. **What that means in practice.** OpenHands does not publish AUD figures, so the numbers below are estimates, converted from typical USD provider costs at roughly AUD 1.40 to the US dollar. GST is generally added by the major model providers, which are registered in Australia. | Usage level | Approx. monthly LLM cost (USD) | Approx. AUD (ex-GST) | | ----------------------- | ------------------------------ | -------------------- | | Light | Under US$10 | \~A$14 | | Regular | US$30–60 | \~A$42–84 | | Heavy (flagship models) | US$150+ | \~A$210+ | These ranges are illustrative, not vendor-published. Your actual cost depends on which model you use and how many tokens you burn. There is no AUD billing. All costs flow in USD, either to your chosen provider or through the OpenHands at-cost provider. Australian businesses should expect GST on those invoices and should keep the receipts for input-tax credits. --- ## How it works in practice You have two front doors. The fast one is OpenHands Cloud: sign in with a GitHub account and start a conversation in the browser. The slower, more private one is self-hosting via Docker, the CLI, or a VM. The workflow is the same once you are in. You describe a task in plain language. The agent reads the relevant files, plans an approach, then writes code, runs commands in a sandboxed terminal, catches errors, and iterates until the task is done or it hits a wall. The agent runs inside an isolated container. The vendor's enterprise page describes a "containerised sandbox runtime" and a Kubernetes-native, cloud-native runtime for the hosted product. For self-hosted deployments, the same page states your code "never leaves your environment" with "no shared tenancy." OpenHands handles genuinely complex, multi-file work. It understands import trees, runs your tests, and fixes failures without being told each step. The integrations matter too: Jira and Slack on the Individual tier, plus GitHub and Notion noted in the docs, let you trigger and track agent work where you already operate. Setup is the honest friction. Cloud is quick. Self-hosting takes real effort — Docker, model keys, sandbox configuration, and resource tuning. A casual user should not expect a five-minute start on the self-host path. The 10-conversations-a-day cap on the free Individual tier is also a real constraint. For a developer using an agent heavily through a working day, that ceiling arrives fast. --- ## What it does well **Cost floor.** The agent is free on both the local and hosted Individual tiers. You pay only for inference, and the vendor states it sells LLM access at cost with no markup. That is a rare, honest pricing posture. **Depth.** This is a full autonomous agent, not autocomplete. Multi-file edits, terminal execution, test-fix loops, parallel agents, scheduling, and an SDK put it among the most capable tools in the category. **Model freedom.** Because it is model-agnostic and open source, you are not locked to one provider. If a model gets too expensive or too slow, you switch. **Self-host control.** The MIT core means you can audit the code and run it on your own infrastructure. For security-conscious teams, that transparency is worth real money. **One control centre.** Running third-party agents such as Claude Code and Codex through one interface, per the vendor docs, is a genuinely useful consolidation for teams juggling several tools. --- ## What it doesn't do well **Setup complexity on self-host.** Getting the self-hosted version running well takes Docker fluency, model configuration, and sandbox tuning. This is not a tool a non-technical operator should attempt to deploy alone. **The free cloud cap.** Ten conversations per day on the Individual tier is tight for daily professional use. Heavy users will feel it within hours and face the jump to Enterprise. **Inference cost is opaque.** Total monthly spend is unpredictable until you have run it for a while. A heavy week on a flagship model can quietly run past A$200. **Enterprise gates the team features.** SSO, SAML, unlimited concurrency, and priority support all sit behind custom Enterprise pricing. Small Australian teams face a binary: do it yourself, or pay for Enterprise. **Free support is community-only.** On the free tiers, support is the community. A priority channel exists only on Enterprise. For business-critical work, that is a gap. --- ## Australian fit OpenHands has no Australian-specific presence. There is no AUD billing, no AU support desk, and no AU-localised documentation. The company is US-based. On data residency, the vendor's public pages do not name a hosting region for OpenHands Cloud. The enterprise page describes a Kubernetes-native cloud runtime but does not state where it runs. I could not verify an Australian (or any specific) cloud region from primary sources, so I will not claim one. That matters under the Privacy Act 1988 (Cth). If your codebase carries personal information about Australians, you must know where it is processed. With OpenHands Cloud, that location is not publicly confirmed, which is a real consideration for regulated workloads. The self-host path is where Australian fit improves. The vendor states that for self-hosted deployments, code never leaves your environment. Run OpenHands Enterprise inside an Australian VPC, point it at an AU-resident model endpoint, and you can keep both code and inference onshore. That is the route privacy-conscious AU teams should take. On compliance, the vendor links a Trust Center but publishes no SOC 2 or ISO 27001 status I could verify on a primary page. I am not asserting a certification the vendor does not clearly state. **Australian Fit score: 2/5.** No AUD billing, no AU support, and no published AU hosting region. Self-hosting can deliver real residency, but only if your team does the configuration work deliberately. > **Note on the rating.** Under our [Methodology](https://www.theaidirectory.com.au/methodology/), a five-star rating or a category Top Pick requires an Australian Fit score of at least 4/5\. OpenHands scores 2/5 here, so it cannot be named a category Top Pick, and its displayed rating is capped at four stars regardless of the raw total. --- ## Alternatives - [**Cline**](https://www.theaidirectory.com.au/review-cline-australia-2026/) — another open-source, BYOK coding agent, focused on VS Code. Lighter to start than self-hosted OpenHands, with the same model-freedom and self-host privacy logic. Compare on whether you want an editor extension or a standalone agent platform. - [**Cursor**](https://www.theaidirectory.com.au/cursor-review-australia-2026/) — the polished proprietary code editor. Faster setup, predictable per-seat USD pricing, but your code goes to its infrastructure. Choose it for plug-and-play simplicity over data control. - [**Claude Code**](https://www.theaidirectory.com.au/claude-code-review-australia-2026/) — Anthropic's terminal-based agent. Deeply capable on reasoning-heavy tasks, billed through Anthropic directly. Worth comparing if you want a single-vendor relationship rather than an open platform. --- ## Bottom line OpenHands is one of the strongest free coding agents available, and among the most capable at any price. The open-source core is real: you can run the whole thing on your own infrastructure and pay nothing for the tool. For Australian teams that care about data control and have the engineering depth to self-host, it is a genuinely good option. The reservations are practical, not philosophical. The free cloud tier is capped low. Self-hosting takes work. And there is no AUD billing, no AU support, and no publicly confirmed Australian hosting region — so the data-residency burden sits with you. Do that work and OpenHands rewards you. Skip it, and a hosted rival like Cursor will be easier to operate. --- ## Score | Criterion | Score | Notes | | --------------------- | --------- | ------------------------------------------------------------------------------------------------------ | | Value for Money | 5/5 | Free agent on local and hosted tiers; LLM sold at cost, no markup. Unmatched cost floor. | | Ease of Use | 2/5 | Cloud is quick, but self-host needs Docker and model setup; free tier capped at 10 conversations/day. | | Feature Depth | 5/5 | Full autonomous agent: multi-file edits, sandboxed execution, parallel agents, SDK, integrations. | | Australian Fit | 2/5 | No AUD billing, no AU support, no published AU region; self-host can deliver residency if configured. | | Reliability & Support | 2/5 | Community-only support on free tiers; SLA and priority support gated to Enterprise; no verified SOC 2. | | **Total** | **16/25** | **⭐⭐⭐⭐ Very Good** | --- Update history - **29 Sep 2026 — verified by browser; tiers unchanged.** openhands.dev/pricing on an Australian connection: Open Source (local) free; Individual (cloud) free, one user, 10 conversations a day, bring your own key or use OpenHands' providers at cost; Enterprise custom — exactly as described below. The page publishes no subscription price in any currency, so the AUD table here stays our own illustrative estimate of model spend. [Source](https://www.openhands.dev/pricing?ref=theaidirectory.com.au) · capture `Dossiers/captures/openhands-pricing-2026-09-29.txt` - **21 Sep 2026** — **The grossed-up AUD column removed from the pricing table.** The table carried a second Australian column adding 10% GST to every converted figure, so each plan had two Australian prices and the higher one asserted a tax most of our readers do not pay: under the ATO rule our GST treatment rests on, a GST-registered Australian business that supplies its ABN is **not charged GST** on an imported digital service. **Nothing was re-cut** — the ex-GST column is unchanged and is the price. Removed, and kept visible here per our price-history rule: the column headed *"Approx. AUD (if GST charged, +10%)"* (Light **\~A$15**; Regular **\~A$46–92**; Heavy (flagship models) **\~A$231+**). The GST position is stated in words once in the note above the pricing table. No USD price moved and no score changed. `reviewed:` untouched per our dating rule. ## Frequently asked questions ### Is OpenHands free to use in Australia? Yes. The open-source local agent is free, and the hosted Individual tier is also free for a single user, capped at 10 conversations per day. You pay only for LLM inference, which OpenHands sells at cost with no markup, or you bring your own provider key. There is no AUD billing — all inference costs are in USD. The Enterprise tier is custom-priced. ### Does OpenHands keep my code in Australia? It depends on how you run it. OpenHands does not publish a hosting region for its cloud, so cloud residency cannot be confirmed. If you self-host OpenHands inside an Australian VPC and point it at an AU-resident model endpoint, the vendor states your code never leaves your environment, keeping both code and inference onshore. For regulated data, the self-host path is the safer route. ### How does OpenHands compare to Cursor for Australian developers? Cursor is a polished proprietary editor with predictable per-seat USD pricing and a fast setup, but your code goes to its infrastructure. OpenHands is free on its core tiers, fully open source, and self-hostable for real data control, but the self-host path takes Docker and model configuration. Choose Cursor for simplicity, OpenHands for cost and data control. --- ## Sources Every primary claim below is paired with a same-day Wayback Machine snapshot captured on 30 June 2026. - OpenHands pricing page: [openhands.dev/pricing](https://www.openhands.dev/pricing?ref=theaidirectory.com.au) — snapshot: [web.archive.org/web/20260630002206](https://web.archive.org/web/20260630002206/https://www.openhands.dev/pricing) (plans, free tiers, 10-conversation cap, at-cost LLM, Enterprise features) - OpenHands homepage: [openhands.dev](https://www.openhands.dev/?ref=theaidirectory.com.au) — snapshot: [web.archive.org/web/20260630002311](https://web.archive.org/web/20260630002311/https://www.openhands.dev/) (product description, open-source claim, deployment models, self-host data claim) - OpenHands GitHub repository: [github.com/OpenHands/OpenHands](https://github.com/OpenHands/OpenHands?ref=theaidirectory.com.au) — snapshot: [web.archive.org/web/20260630002409](https://web.archive.org/web/20260630002409/https://github.com/OpenHands/OpenHands) (78,000+ stars, agent capabilities, self-hosting) - OpenHands licence (MIT, with separate enterprise/ licence): [github.com/OpenHands/OpenHands/blob/main/LICENSE](https://github.com/OpenHands/OpenHands/blob/main/LICENSE?ref=theaidirectory.com.au) — snapshot: [web.archive.org/web/20251116105708](https://web.archive.org/web/20251116105708/https://github.com/OpenHands/OpenHands/blob/main/LICENSE) (MIT core; enterprise/ dual-licensed) - OpenHands Enterprise page: [openhands.dev/enterprise](https://www.openhands.dev/enterprise?ref=theaidirectory.com.au) — snapshot: [web.archive.org/web/20260620184142](https://web.archive.org/web/20260620184142/https://www.openhands.dev/enterprise) (containerised sandbox, Kubernetes-native runtime, self-host "code never leaves your environment") --- --- *The AI Directory Australia is an independent editorial publication. We don't accept payment for positive reviews. Read our* [*Methodology*](https://www.theaidirectory.com.au/methodology/) *and* [*Affiliate Disclosure*](https://www.theaidirectory.com.au/affiliate-disclosure/)*.*