Tabnine
Private, self-hostable AI coding assistant, now part of Tricentis
Overview
Tabnine is an enterprise AI coding assistant and agent that can run as SaaS, in a VPC, on premises or fully air-gapped, with no storage of or training on customer code. Tricentis acquired it in July 2026 and tabnine.com now redirects to Tricentis, so plans are quoted through Tricentis sales. The last published prices were $39 and $59 per user per month on annual subscriptions.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 14 Oct 2026 and may be out of date or incomplete.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.
- Current-line and multi-line code completions
- IDE chat using models from Anthropic, OpenAI, Google, Meta, Mistral and others
- Jira Cloud and Data Center integration
- SaaS, VPC, on-premises or fully air-gapped deployment
- Zero code retention; IP indemnification subject to terms
- Autonomous agents with optional user-in-the-loop oversight
- Tabnine CLI for terminal, remote and CI use
- MCP tools for Git, Jira, Confluence, Docker and CI/CD
- Tabnine Context Engine with unlimited Bitbucket, GitHub, GitLab and Perforce connections
- Per-user and per-team spending thresholds
- Agents run remotely or headless in CI/CD pipelines
Use cases
What you can produce with Tabnine
- Code completions and IDE chat in VS Code, Cursor, JetBrains IDEs, Eclipse and Visual Studio
- Deployment as SaaS, in a VPC, on premises or fully air-gapped
- No storage of, or training on, customer code
- Agents grounded by the Context Engine in repositories, Jira and Confluence
- An OpenCode-based Tabnine CLI for terminal and CI workflows, with read access to Jira issues
- Admin analytics, model access controls and per-team spending limits
Tabnine's case is deployment control: air-gapped, VPC or on-premises installation, no retention of customer code, models trained on permissively licensed code, and a Context Engine that grounds agents in an organisation's repositories and tickets. The product is still shipping, with an OpenCode-based CLI released to all customers in September 2026.
The catch is ownership. Tricentis bought Tabnine for its Context Engine and now routes the website and pricing to its own sales team, so long-term plans for the coding assistant are not public. Worth shortlisting for regulated teams that need self-hosting; ask Tricentis for a written roadmap before signing a multi-year deal.
What we found
Tabnine is one of the oldest AI coding assistants, and in 2026 it stopped being a standalone company. On 30 July 2026 Tricentis, the Austin-based software-testing group, announced that it had acquired Tabnine in order to build Tabnine's Enterprise Context Engine into the Tricentis Agentic Quality Engineering Platform. Terms were not disclosed. Calcalist estimated the deal at tens of millions of dollars, and Shibolet, the law firm that acted for Tabnine, describes the consideration as tens of millions of US dollars paid in Tricentis shares and retention arrangements. Calcalist also reports that Tabnine had raised $55 million and cut 15 of its 80 staff the year before.
The public website has since been folded away. By late September 2026 tabnine.com, its pricing page and its blog all redirected to tricentis.com, with pricing landing on a Tricentis contact form. The last pricing page captured before that, on 18 September 2026, listed two plans on annual subscriptions: the Code Assistant Platform at $39 per user per month, with completions, IDE chat over models from Anthropic, OpenAI, Google, Meta and Mistral, and Jira integration; and the Agentic Platform at $59, adding autonomous agents, the Tabnine CLI, MCP tools and the Context Engine. Using Tabnine-hosted models costs extra, at the provider's price plus a 5% handling fee. No free or individual plan was listed.
The product is still shipping. The documentation remains live, and the private-installation release notes run to version 6.6.9 on 8 October 2026: 6.6.0 made an OpenCode-based Tabnine CLI available to all customers, and 6.6.9 let it read Jira issues. Calcalist reports that existing customers will keep receiving support.
The selling point is unchanged: SaaS, VPC, on-premises or fully air-gapped deployment, no storage of or training on customer code, and models trained only on permissively licensed open-source code. For ASEAN banks and public-sector teams with strict data-residency rules it remains credible, but buying now means a conversation with Tricentis sales, and the roadmap belongs to a testing company.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
Notable facts
- Tabnine was originally called Codota before rebranding in 2020, and was one of the first AI tools to complete entire function bodies rather than just single lines.
- Tabnine's self-hosted Enterprise model runs the full AI inference on-premises — code never leaves your servers, not even for model updates.
- The tool was trained on permissively licensed code only, meaning it avoids training on GPL-licensed repositories that could create licence compliance issues.
Frequently asked questions
About this listing
This entry was compiled from publicly available data including Tabnine's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Tabnine 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 Tabnine directly →
Spotted something out of date? Suggest an update →
Alternatives to Tabnine
More in Code & Dev Tools
