> ## 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. # Tabby Review: The Self-Hosted AI Coding Assistant (Australian Edition) - URL: https://www.theaidirectory.com.au/review-tabby-australia-2026/ - Published: 2026-06-25T23:00:00.000Z - Updated: 2026-10-09T06:36:38.000Z - Description: Tabby is an open-source AI coding assistant you run on your own server — no code leaves your infrastructure unless you choose it to. Here is what that privacy-first architecture means for Australian development teams. - Author: Adam - Tags: Coding & Development, AI Reviews, #review **15**/25 ★★★☆☆ Good TabbyML, Inc. Verified 9 Oct 2026 Reviewed 26 Jun 2026 Value for Money**4**/5 Ease of Use**2**/5 Feature Depth**3**/5 Australian Fit**4**/5 Reliability & Support**2**/5 From **Free** · paid tiers on request **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):** Tabby is the most privacy-complete AI coding assistant available. Run it on your own server and code never touches an external vendor at all. That is a genuinely rare property and a meaningful advantage for Australian teams handling sensitive IP or regulated data. The trade-off: setup demands real infrastructure work, you need a capable GPU to get good performance, and the product experience is a long way behind polished hosted alternatives. For security-first teams with the resources to run it, Tabby earns its place. For everyone else, the setup overhead will be too high. > > **Score: 15/25 — ⭐⭐⭐ Good** > > *Last reviewed: 26 June 2026 by Adam.* --- ## What Tabby is Tabby is an open-source, self-hosted AI coding assistant built by TabbyML, Inc. The GitHub repository has 21,600+ stars and the project is licensed under the Apache 2.0 licence. The tagline is accurate: it is a self-contained alternative to GitHub Copilot that runs entirely on your own infrastructure. Unlike Cline, which is an agent layer that connects to external model providers, Tabby runs the model itself. You download Tabby, point it at your GPU (or a CPU-only server if you are patient), configure a supported model, and run a server that your developers' IDE plugins connect to. Code stays on your network. The LLM processes requests locally. The feature set covers the main coding assistant use cases: real-time code completion, an answer engine (chat about your codebase), inline chat directly in the editor, and data connectors that let Tabby index your repositories for context-aware suggestions. As of June 2026, TabbyML has added "Pochi" — a full-stack AI teammate feature on the cloud path — though the core self-hosted offering remains the product's main identity. Tabby supports VS Code, Neovim, IntelliJ IDEA, PyCharm, WebStorm, Android Studio, GoLand, PhpStorm, Rider, RubyMine, CLion, AppCode, and Eclipse — one of the widest IDE coverage lists of any AI coding assistant. Compatible models include CodeLlama, StarCoder, Qwen2.5-Coder, DeepSeek Coder, and others from the HuggingFace ecosystem. You choose the model that fits your hardware. --- ## What it costs in Australia Tabby's pricing structure has two paths: self-hosted and cloud. **Self-hosted pricing:** | Plan | Cost | Users | Notes | | ---------- | ---------------- | --------- | ------------------------------------------------------------------------------ | | Community | Free | Up to 5 | Full self-hosted deployment; code completion, answer engine, inline chat | | Team | US$19/seat/month | Up to 50 | All Community features + team management, usage analytics, flexible deployment | | Enterprise | Custom (contact) | Unlimited | Enhanced security, SSO, custom deployment, dedicated Slack support | There is no AUD billing. All pricing is in USD. At approximately AUD 1.40 to the USD, plus GST where applicable: - Team plan: \~A$27–32/seat/month (ex-GST \~A$27; ex-GST \~A$27) - Community: Free **Cloud path (Tabby Cloud / Pochi):** Tabby's cloud product offers US$20 in free monthly credits. Beyond that, costs are token-based, billed on LLM usage. Tab completion remains free on the cloud path with no usage limits. (specific model pricing tiers not confirmed from primary source as of review date) For most Australian teams evaluating Tabby, the self-hosted Community or Team plan is the relevant path. **Sources:** Tabby pricing page (tabbyml.com/pricing, accessed 26 June 2026). --- ## How it works in practice Setup is the first test. You need a server — ideally with a GPU — and Docker or a direct binary installation. The Tabby documentation walks through the process clearly, but expect to spend at minimum a few hours getting it running properly. For a small team, a single GPU server (NVIDIA RTX series or better) is sufficient for CodeLlama or Qwen2.5-Coder serving a handful of concurrent developers. Once running, developers install the Tabby IDE plugin, configure the server URL, and code completions start appearing inline. The experience from there feels familiar — Tab completes code, the answer engine responds to questions about the codebase, and inline chat lets you refactor or explain code without leaving the editor. The repository indexing feature (Context Providers) is worth configuring. Tabby can index your codebase and use it as additional context for suggestions, making completions more project-specific and less generic. This takes some configuration but noticeably improves suggestion quality on proprietary codebases. Suggestion latency depends almost entirely on your hardware. A well-configured GPU server returns completions in under a second. A CPU-only setup on a small cloud instance can be noticeably slow. If you are evaluating Tabby for a team, hardware investment is a meaningful part of the total cost calculation. --- ## What it does well **True data sovereignty.** No other AI coding assistant with comparable IDE coverage gives you this level of control. When deployed on-premises or on an Australian cloud server (AWS Sydney, Azure Australia East, or your own hardware), code never leaves your jurisdiction. This is the product's defining property. **Broad IDE support.** VS Code, JetBrains (all major IDEs), Neovim, Eclipse, and more. Very few AI coding tools cover this much ground. **Model flexibility.** You choose the LLM. You can run efficient smaller models for speed and cost, or more capable models if your hardware supports it. You can fine-tune on your own repositories. **No per-seat cost on Community.** For teams of up to five developers, the tool is genuinely free — not a time-limited trial. The only cost is infrastructure. **Active open-source community.** 21,600+ GitHub stars, regular releases (v0.32.0 in January 2026), and a Slack community. The project is maintained and improving. --- ## What it doesn't do well **Setup demands real infrastructure investment.** A capable GPU server running Tabby costs real money — a cloud GPU instance in AWS Sydney capable of serving a small team could easily run A$500–1,500/month depending on instance type. (estimate) This makes the "free" Community plan significantly less free in practice. Factor the infrastructure cost in before assuming Tabby is cheaper than GitHub Copilot. **Suggestion quality varies by model.** The best suggestions require the largest models, which require the most GPU memory. On constrained hardware, the quality delta versus hosted products like Copilot or Cursor is noticeable. **No autonomous agent mode.** Tabby is an IDE assistant: code completion, chat, inline editing. It does not autonomously plan and execute multi-file changes the way Cline or Cursor Composer do. For teams wanting an agent rather than an assistant, this is a meaningful gap. **Enterprise features require paid tiers or custom contracts.** SSO and advanced security controls are not in the free Community tier. **Limited AU support.** TabbyML is a US-based company (TabbyML, Inc.). Support for the paid Team plan is email-based; Enterprise customers get a dedicated Slack channel. There is no Australian-specific support. --- ## Australian fit Tabby's self-hosting architecture is the strongest Australian-fit story of any AI coding assistant we have reviewed. Deploy it on an AWS `ap-southeast-2` (Sydney) instance, an Azure Australia East server, or your own on-premises hardware, and code stays in Australia. Full stop. This matters in two specific contexts: 1. **Privacy Act 1988 compliance.** If your codebase contains personal information about Australians — client data embedded in source code, test data, configuration files — self-hosting means that data stays under your control and subject to Australian jurisdiction. 2. **IP protection.** For Australian startups and agencies with commercially sensitive code, sending code to any US-based API introduces export and IP risk. Tabby eliminates that risk entirely when self-hosted. For teams on the cloud path using Pochi, code will be processed by TabbyML's cloud infrastructure (US-based). The data residency advantage disappears unless you negotiate specific terms. The pricing gap in AUD terms: the Team plan at \~A$29/seat/month is competitive with GitHub Copilot Business (approximately A$22/seat/month). When you add infrastructure costs for self-hosting, the economics depend on team size and hardware efficiency. Large teams amortise infrastructure costs across more developers; small teams may find hosted alternatives cheaper overall. Tabby has no AUD billing. It does not publish AU-specific pricing or GST information. We recommend confirming GST treatment on the Team plan subscription directly with TabbyML before purchase. **Australian Fit score: 4/5.** Self-hosting gives genuine, configurable data sovereignty — the best AU-fit story in this category. Loses one point for no AUD billing, no AU-specific support, and infrastructure costs that reduce the free-tier value for small teams. --- ## Alternatives - [**Cline**](https://www.theaidirectory.com.au/review-cline-australia-2026/) — another open-source coding tool, but operates as an autonomous agent rather than an assistant. Routes to external model providers by default, though supports local Ollama models. Better for complex agentic tasks; Tabby is better for team-wide inline completion with full data control. - [**Continue**](https://www.theaidirectory.com.au/review-continue-australia-2026/) *(wound down — the hosted product ended 15 July 2026; the open-source code remains)* — open-source, BYOK coding assistant with VS Code and JetBrains support. Also supports local models via Ollama. Lighter weight than Tabby; does not run its own server layer. - [**GitHub Copilot**](https://www.theaidirectory.com.au/github-copilot-review-australia-2026/) — the market-leading hosted coding assistant. Better polished experience; no self-hosting option. Code is sent to GitHub/OpenAI infrastructure. --- ## Bottom line Tabby is the right tool for a specific Australian buyer: a development team that needs AI coding assistance but cannot send code to external servers — for IP, regulatory, or security reasons. The self-hosted architecture is genuinely complete and the feature set covers the real-world use cases. If that is your situation, Tabby is worth the setup effort. If you do not have a pressing data residency requirement, the infrastructure overhead and experience gap versus hosted alternatives make it a harder sell. The Community tier is excellent for small teams trialling it; the Team tier pricing is reasonable for larger groups once you account for infrastructure costs. --- ## Score | Criterion | Score | Notes | | --------------------- | --------- | ------------------------------------------------------------------------------------------------------------------------------------------------- | | Value for Money | 4/5 | Free Community tier for up to 5 users; Team plan competitive at \~A$27/seat. Must factor in GPU/server infrastructure cost. | | Ease of Use | 2/5 | Real setup effort required: Docker, GPU server, model configuration. Not plug-and-play. | | Feature Depth | 3/5 | Solid inline completion and chat; no autonomous agent mode. IDE coverage is the widest in the category. | | Australian Fit | 4/5 | Self-hosting gives genuine AU data sovereignty. No AUD billing or AU-specific support, but data residency story is the best in class. | | Reliability & Support | 2/5 | Community support on free tier; email on Team; dedicated Slack on Enterprise. Performance depends on your hardware. No uptime SLA on lower tiers. | | **Total** | **15/25** | **⭐⭐⭐ Good** | --- Update history - **9 Oct 2026** — Re-checked for the AUD-claim queue (#470), tabbyml.com/pricing read on an Australian connection: Community free and open source for up to 5 users, Team US$19 per seat a month for up to 50 users, Enterprise on request, and the cloud product's US$20 of free monthly credits with tab completion always free — the table stands, bare `$`, no AUD, so `aud_native: false` stands. Capture: `Dossiers/captures/tabby-pricing-2026-10-09.txt`. ## Frequently asked questions ### Can I run Tabby in Australia so my code never leaves the country? Yes. Tabby is self-hosted, which means you run it on your own server. Deploy it on an AWS Sydney (ap-southeast-2) instance, an Azure Australia East server, or your own on-premises hardware, and code stays in Australia. No external AI vendor processes your code. This is the tool's primary advantage for Australian teams with data sovereignty requirements. ### What does Tabby cost in Australia? Tabby's Community plan is free for up to 5 users but requires you to provide your own server infrastructure (GPU recommended). The Team plan is US$19/seat/month — approximately A$27/seat/month ex-GST — for up to 50 users. Enterprise pricing is custom. There is no AUD billing; all plans are priced in USD. ### Does Tabby work with JetBrains IDEs? Yes. Tabby has plugins for all major JetBrains IDEs including IntelliJ IDEA, PyCharm, WebStorm, GoLand, PhpStorm, Rider, CLion, RubyMine, and Android Studio. It also supports VS Code, Neovim, and Eclipse — one of the widest IDE coverage sets of any AI coding assistant. --- ## Sources - Tabby official website: [tabbyml.com](https://www.tabbyml.com/?ref=theaidirectory.com.au) (accessed 26 June 2026) - Tabby pricing page: [tabbyml.com/pricing](https://www.tabbyml.com/pricing?ref=theaidirectory.com.au) (accessed 26 June 2026) - Tabby documentation: [tabby.tabbyml.com/docs/welcome](https://tabby.tabbyml.com/docs/welcome/?ref=theaidirectory.com.au) (accessed 26 June 2026) — [Wayback 24 May 2026](https://web.archive.org/web/20260524005408/https://tabby.tabbyml.com/docs/welcome/) - Tabby GitHub repository: [github.com/TabbyML/tabby](https://github.com/TabbyML/tabby?ref=theaidirectory.com.au) (accessed 26 June 2026) — [Wayback 19 June 2026](https://web.archive.org/web/20260619060405/https://github.com/TabbyML/tabby) --- --- *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/)*.*