Description
Tabnine is an AI coding agent built for engineering teams that need code assistance to stay inside their own infrastructure. It connects directly to an organization's repositories and development systems, and is designed so no code has to leave that environment to get AI help.
Read More
The platform runs on a Kubernetes cluster, either on-premises or on a supported cloud provider. Three pieces make it up: the Tabnine client, an IDE plugin developers install locally; the Tabnine cluster, which runs the inference service on GPU-accelerated nodes plus a data plane for analytics (no code is stored, and user identifiers are masked) and a control plane for identity and configuration; and the AI models themselves.
Tabnine Chat gives on-demand answers about a codebase: explaining code, generating snippets, or answering documentation questions. Tabnine Agent goes further. It is an autonomous, task-oriented assistant that can carry out codebase-wide refactoring, generate tests, write documentation, and check code against policy, while keeping a tight feedback loop with the developer and checking in for approval before proceeding on complex work. Both are also available through a command-line interface, alongside inline code completions.
Tabnine trains two kinds of proprietary models: universal models trained on public, permissive open-source code, available to all users, and fine-tuned models trained further on a customer's own codebase, available only on the Enterprise plan. Pro users can also switch Chat to selected third-party models under those providers' own terms. Deployment can run as Tabnine's hosted SaaS, the default for both Pro and Enterprise, or, for some Enterprise customers, as a private installation inside the customer's own VPC or fully on-premises, where only Tabnine's own protected model is available.
Tabnine was recently acquired by Tricentis, the enterprise software testing company, which is extending the coding agent into its own quality-engineering platform. Tabnine's documentation and release notes continue to be maintained on their own docs site.
- CategoryAI Coding Assistants
- Pricing ModelPaid
- DeploymentCloud, On-premises
- PlatformsWeb
Tabnine's Core Features
Inline code completions in supported IDEs
Tabnine Chat for on-demand answers, explanations, and generated snippets about your codebase
Tabnine Agent for autonomous multi-step tasks like refactoring, test generation, and documentation, with developer check-ins on complex work
Tabnine CLI for agent and chat workflows outside the IDE
Universal AI models trained on public, permissive open-source code
Fine-tuned models trained further on a customer's own codebase (Enterprise only)
SaaS hosting by default, with private installation in a customer's own VPC or on-premises for some Enterprise customers
A data plane that stores no code and masks user identifiers
How to use Tabnine?
Install: add the Tabnine plugin to your IDE from its listing in Tabnine's docs.
Get completions: start typing and accept inline suggestions as you code.
Ask Tabnine Chat: open Chat to ask questions about your codebase or generate a snippet.
Run Tabnine Agent: hand it a larger task, like a refactor or test suite, and approve its plan when it checks in.
Use the CLI: run Tabnine Agent or Chat from the command line for the same workflows.
Choose a deployment: stay on Tabnine's SaaS cloud, or ask about private installation for an air-gapped or on-prem setup.
Tabnine's Use Cases
- Enterprise code completion
- Autonomous refactoring
- Air-gapped deployment
- Codebase Q&A
