Tabnine

Tabnine's stated selling point is that it can run as SaaS, in your VPC, fully on-premises, or fully air-gapped, with zero code retention, no training on customer code, and no sharing with third parties — the same privacy posture is described identically across both plan tiers on the pricing page. This is a real architectural distinction from most AI coding assistants, which are SaaS-only. It also supports bringing your own LLM endpoint (on-prem or cloud) for unlimited usage, or Tabnine-hosted models billed as reserved token consumption at provider price plus a 5% handling fee.

As of today, Tabnine sells two seat-based plans, both billed annually and both routed through a 'Get a quote' contact-sales flow rather than self-serve checkout: Code Assistant at $39/user/month (completions, IDE chat, IP indemnification, standard deployment/security features) and the Agentic Platform at $59/user/month (adds autonomous coding agents, a CLI agent, MCP tool support, and the Context Engine that indexes your codebase, Jira, and Confluence). A separate capacity-based product, Headless Agents, prices CI/CD automation by monthly token-processing capacity rather than seats: $1,200/month for up to 5B tokens (Business) or $5,000/month for up to 50B tokens (Enterprise) — customers pay their own LLM provider costs on top in both cases. There is no published free tier and no self-serve individual plan; every tier requires talking to sales.

Tabnine previously offered free and low-cost individual plans; it discontinued the free Basic tier and, more recently, retired the ~$9-12/month individual Dev/Pro plans to go enterprise-only, per third-party coverage — we could not verify the exact retirement date from a Tabnine-published source. What we could verify directly: Tabnine's Trustpilot page currently shows a 'Poor' rating of 2.1/5 from just 10 reviews, and the most recent ones (as late as June 2026) are specifically former individual/Pro subscribers describing account cancellations, broken login flows, and the switch to enterprise-only pricing. That's a small sample, but it's a consistent, on-topic complaint pattern rather than generic negativity, and it corroborates the broader trend independently.

On Gartner Peer Insights, in the Enterprise AI Coding Agents category, Tabnine holds a 4.5/5 average across 14 ratings (57% five-star, 21% four-star, 21% three-star, none below three stars) — a much more positive picture than Trustpilot's consumer reviews. Both samples are small (10 and 14 reviews respectively), so neither should be read as definitive, but the split itself is informative: it lines up with a company that has explicitly repositioned around enterprise buyers and away from individual developers.

Tabnine's IDE integrations (VS Code, JetBrains, Vim, and others) are open-source under the codota/tabnine GitHub organizations — the VS Code extension (github.com/codota/tabnine-vscode) is MIT-licensed, actively maintained (last push within the past two weeks as of this writing), with 1,435 GitHub stars and no archived status. The completion/agent backend itself is closed source. No CVEs or advisories for Tabnine were found in the GitHub Advisory or OSV.dev databases at time of writing.

Tabnine says it has operated since 2018 (the company was originally Codota). Its most recent disclosed round was a $25M Series B in November 2023, led by Telstra Ventures with Atlassian participating as a new investor — confirmed directly from Tabnine's own blog post. Third-party trackers (Tracxn, Crunchbase-derived aggregators) report total funding around $100M+ across five rounds, but we could not verify that cumulative figure against a primary Tabnine source, so treat it as a third-party estimate, not a confirmed number.

Tabnine runs a public trust center (trust.tabnine.com, via SafeBase) that lists SOC 2 and ISO/IEC 27001 certifications, with GDPR compliance also referenced on the pricing page and docs. Detailed audit reports are gated behind an access request, which is standard practice for this kind of trust portal, not unusual for the category.