Tabnine

Enterprise AI coding and agent platform

by Tabnine

Overview

Tabnine is an enterprise coding platform in two parts: a Code Assistant for completion and chat inside the IDE, and an Agentic Platform that adds agents, a CLI, MCP and a Context Engine, with the two separated by plan rather than both included everywhere. Deployment is a choice rather than a property: SaaS, private or fully isolated, where SaaS inference sends code context to Tabnine's cluster, ordinary self-hosting keeps the model in your environment but still reports operational logs and metrics to Tabnine with a stated one-week retention, and fully isolated deployments can use the customer's own logging and Prometheus instead - so zero code retention is not the same as nothing being transmitted, and operational telemetry is not the code sent for inference.

The models are layered as well: Tabnine documents training its completion and Protected chat models on permissively licensed code, which does not extend to third-party models a customer selects, and a custom model trained on a customer's code inside their environment is a separate arrangement again. The current public pricing lists Code Assistant at 39 per user a month and Agentic Platform at 59 per user a month, both as annual subscriptions, in a currency the page did not show us; when Tabnine supplies LLM access, model provider token prices apply plus a 5% fee, and platform use is unlimited when you bring your own model or endpoint, which does not make inference free.

Supported editors are VS Code, JetBrains IDEs on the IntelliJ platform, Eclipse and Visual Studio 2022/2026 on Windows, without a guarantee that every feature matches across them. The current starting price is unverified: the pricing page returned HTTP 502 with an empty body on 12 October 2026, as it had on 10 October 2026, and the only text readable that day was the same URL's Wayback snapshot dated 18 September 2026, which again printed the per-user amounts with no currency beside them, so treat the amount as a quote to confirm rather than a figure to compare.

Editor's take

An enterprise option when you need to control where inference happens - but that is decided deployment by deployment rather than by the brand. Three things to settle first: which of the two products you need, because the agentic features sit on the higher plan; whether your deployment still transports code for inference and reports operational telemetry, because zero code retention is not zero transmission; and what the model endpoint costs, since bringing your own model makes the platform unlimited while the inference is not free. The pricing page we could read stated per-user amounts without a currency and gave no verified monthly option, so treat the price as a quote.

Our reading of who this suits, based on the sources above. Opinion, offered to help you decide - not a lab result.

Key Features

Code Assistant: completion and chat in the IDE
Agentic Platform: agents, CLI, MCP and Context Engine on the higher plan
Deployment choices: SaaS, private or fully isolated, with different telemetry behaviour
Completion and Protected chat models documented as trained on permissively licensed code
Bring your own model or cloud endpoint for unlimited platform use; a 5% fee applies on provider token pricing when Tabnine supplies access
Supported IDEs: VS Code, JetBrains on the IntelliJ platform, Eclipse, and Visual Studio 2022/2026 on Windows

Use Cases

Completion and chat in an IDE you already use
Regulated environments that need private or isolated deployment
Teams that want a CLI and agents in the same platform
Organisations bringing their own model endpoint

Pros

✓Deployment is a documented decision rather than a fixed SaaS model
✓The assistant and agentic products are described separately instead of merged
✓Bringing your own model or endpoint removes the platform limits
✓The supported editors and the training scope of Tabnine's own models are stated

Cons

✕The page we read offers no free tier; the public entry is an annual per-user subscription
✕The agentic features are not in the lower plan
✕SaaS inference sends code context to the cluster, and standard self-hosting still reports operational logs and metrics to Tabnine with a stated one-week retention - zero code retention is not zero transmission
✕Training on permissively licensed code covers Tabnine's completion and Protected chat models, not third-party models a customer selects
✕The pricing page states 39 and 59 per user a month without a currency we could read, and no monthly option, seat minimum or token bundle we could verify
✕Private deployment means running infrastructure, and model cost follows the endpoint and the contract

Frequently Asked Questions

Is there a free plan?

The current public entry we could read describes annual per-user subscriptions for the Code Assistant and the Agentic Platform rather than a free tier, so this site no longer lists it as free. That is not evidence that historic free accounts were withdrawn - only that the page we read does not offer one.

Does self-hosting keep my code inside my environment?

It depends on the deployment. SaaS inference sends code context to Tabnine's cluster; standard self-hosting keeps the model in your environment but still reports operational logs and metrics with a stated one-week retention, while a fully isolated deployment can use your own logging and Prometheus. Zero code retention is therefore not the same as nothing being transmitted, and operational logs are not the code sent for inference.

What does it cost?

The page we read lists Code Assistant at 39 per user a month and Agentic Platform at 59 per user a month, both annual subscriptions, but it did not show a currency we could verify, so treat those numbers as needing confirmation rather than as dollars. When Tabnine supplies LLM access, provider token prices apply plus a 5% fee; bringing your own model or endpoint makes platform use unlimited, which does not make inference free.

What is my code used to train?

Tabnine documents using permissively licensed code to train its completion and Protected chat models. That does not extend to third-party models you select, and a custom model trained on your code inside your own environment is a separate arrangement.

Task guides for this product

T7 · Check whether code and prompts stay on your Mac

Key facts and where they come from

PriceNot established: The live pricing page could not be read in this round (Internal Error; origin HTTP status unavailable). Current currency, regional purchase eligibility and billing commitment remain unconfirmed; the older 39/59 annual-subscription record is not a live quote.Source: Tabnine - pricing · tabnine official tab-plan
AvailabilityNot established: Official documentation lists supported IDEs and SaaS/private deployment routes, but not current country-by-country distribution or eligibility. No network reachability test was performed.Source: Tabnine - deployment options · Tabnine - supported IDEs
Commercial useNot established: The output-use terms could not be read in this round (Internal Error; origin HTTP status unavailable). Private deployment and training-data licensing do not establish commercial-use permission for generated code.Source: Tabnine - privacy documentation · Tabnine - deployment options
Training on your dataDocumented: Historical scope read on 2026-09-23, not rechecked in full on 2026-10-15: training on permissively licensed code covers Tabnine's completion and Protected chat models, not third-party models a customer selects.Source: Tabnine - privacy documentation
API billingNot established: The current pricing body remains unreadable. The older provider-token-plus-5% record was not reverified, and current provider/API charges or subscription inclusion remain unconfirmed.Source: Tabnine - pricing · tabnine official tab-plan
Export formatsNot established: The official supported-IDE table establishes client compatibility, but not downloadable generated-code artefact or conversation-export file types and their plan boundaries.Source: Tabnine - supported IDEs
Usage limitsNot established: The Enterprise SaaS help and failed live-pricing read do not establish current per-plan minimum seats, token bundles or request quotas. The older BYOM unlimited-platform wording is not treated as free inference.Source: Tabnine - pricing · tabnine official tab-plan

Official sources

The vendor pages we read for the claims on this page, what each one states, and the day we read it.

Tabnine - pricing ↗
Read on 23 September 2026, and re-checked on 12 October 2026. Code Assistant is listed at 39 per user a month and Agentic Platform at 59 per user a month, both as annual subscriptions; the page shows no currency symbol or code, and no monthly option, seat minimum or token bundle is stated. Where Tabnine supplies LLM access, provider token prices apply plus a 5% fee, while bringing your own model or endpoint makes platform use unlimited without making inference free. The live URL returned HTTP 502 with an empty body on 10 October 2026 and again on 12 October 2026, so the only text available on 12 October 2026 was this same URL's Wayback snapshot dated 18 September 2026: it prints "39" against "per user per month" and "{ Annual subscription * }", and "59 per user per month { Annual subscription * }", with no currency beside either amount. The currency therefore remains unestablished and the entry publishes no starting price. The same archived page carries the notice "Tabnine has been acquired by Tricentis", and the www.tabnine.com home page now redirects to tricentis.com/products; that is a vendor notice, not a price.
Read on October 12, 2026
Tabnine - privacy documentation ↗
Historical registered scope from 2026-09-23; the entire statement was not re-read on 2026-10-15: SaaS inference sends code context to the Tabnine cluster, while standard self-hosted deployments still report operational logs and metrics with a stated one-week retention; training on permissively licensed code covers Tabnine's completion and Protected chat models, not third-party models a customer selects. Official body read on 2026-10-15, limited to this scope: The client sends some local project code/context to the Tabnine cluster for inference; local-retention claims do not mean inference stays on the Mac. The page separately describes logs/metrics in self-hosted and air-gapped deployments. Historical scope read on 2026-09-23, not rechecked in full on 2026-10-15: training on permissively licensed code covers Tabnine's completion and Protected chat models, not third-party models a customer selects.
Read on October 15, 2026
Tabnine - deployment options ↗
The official page documents Secure SaaS and private installations in VPC/on-premises, with air-gapped options. Deployment location is not a declaration of country-by-country access. Historical registered scope read on 2026-10-15, not rechecked in full on 2026-10-15: Historical registered scope from 2026-09-23; the entire statement was not re-read on 2026-10-15: Deployment ranges from SaaS to private and fully isolated, and fully isolated deployments can use the customer's own logging and Prometheus instead of reporting to Tabnine - so zero code retention does not mean nothing is transmitted. Official body read on 2026-10-15, limited to this scope: SaaS is hosted; private installations are Kubernetes clusters in a customer VPC or customer servers, with air-gapped Enterprise available. This does not establish a supported single-Mac model endpoint/telemetry setup under$20.
Read on October 15, 2026
Tabnine - supported IDEs ↗
The official table lists VS Code, JetBrains, Eclipse and Visual Studio, with Windows/macOS/Linux support varying by IDE; it does not publish a country-distribution policy or export artefact-format matrix. Historical registered scope read on 2026-09-23, not rechecked in full on 2026-10-15: Supported editors are VS Code, JetBrains IDEs on the IntelliJ platform, Eclipse and Visual Studio 2022/2026 on Windows, with no guarantee that individual features match across them.
Read on October 15, 2026
Tabnine - code privacy page (scope conflict) ↗
This page's statement that no third-party API is used is scoped to Tabnine's own models and does not override the current pricing page, which allows other LLM providers - the two are about different configurations.
Read on September 23, 2026
tabnine official tab-plan ↗
The Enterprise SaaS page describes the enterprise subscription, but the readable body does not settle current per-plan token bundles, request quotas and live provider fee commitments.
Read on October 15, 2026

Compare with other tools in Code Tools

Product	Maker	Pricing	Launched	Shared tags
Claude Code	Anthropic	Paid	2025-02-24	Coding
Devin Desktop	Cognition	Freemium	2024-11-13	Coding
Replit Agent	Replit	Freemium	2024-09	Coding
Cursor	Cursor	Freemium	2023-03	Coding
GitHub Copilot	GitHub	Freemium	2021-10	Coding
Lovable	Lovable	Freemium	2024-12	—