A coding agent for IntelliJ IDEA, powered by local LLMs

Your code never leaves your machine.

Private AI is a coding agent that runs entirely on models you host yourself, through Ollama. It explains, refactors and writes tests for your code, and no prompt, file or keystroke is ever sent to a cloud service.

Start your 30-day free trial See where your data goes
  • No cloud APIs or accounts
  • No telemetry
  • Works fully offline
Private AI Chat
YouWrite unit tests for the discount tiers.
AssistantHere are boundary tests for total() covering 1,000 and 10,000, plus an empty list:
Ollama · localhost:11434 qwen2.5-coder Network: local only
Privacy by architecture

Where does your code go when you ask an AI for help?

With most AI coding assistants, the snippet you highlight is sent over the internet to someone else's servers. With Private AI, the whole conversation stays on hardware you control. This comes from how the plugin is built, not from a policy you have to trust.

Cloud assistant

Code leaves your network

Prompts and file context are sent to a third party and handled under their retention terms.

Private AI

Code stays on your machine

The plugin talks only to the Ollama address you configure. By default that is localhost.

You can also point Private AI at an Ollama server on your company network. Your code then goes to that server, which you control, and still never to us or anyone else.

Private by default, not by setting

Built for code that can't leave the building

Zero data egress

Prompts, selected code and responses only travel between your IDE and your Ollama instance. We never receive them.

No telemetry

No usage analytics, crash reporting or "anonymous" statistics. The plugin makes no network calls except to Ollama.

Air-gap ready

Once a model is pulled, everything works with the network cable unplugged. License keys are also checked offline.

Your models, your rules

Choose any model Ollama can run. Swap, pin or fine-tune it, with no vendor lock-in and no surprise model changes.

Features

An AI pair programmer that lives in your IDE

Private AI fits into IntelliJ IDEA the way the IDE's own tools do: a tool window for conversation and right-click actions for the code in front of you.

Conversational chat

A dedicated tool window with chat history, a multiline composer that grows as you type, and answers formatted with code blocks.

Enter sends · Shift+Enter adds a new line

Explain Code

Select unfamiliar or legacy code, right-click, and get a plain-language walkthrough of what it does and why.

Editor › Private AI › Explain Code

Refactor / Optimise

Suggestions that improve readability and performance while keeping the behaviour of the selected code the same.

Editor › Private AI › Refactor / Optimise

Generate Unit Tests

Draft tests for the active file or the current selection, covering edge cases and boundaries you might otherwise miss.

Editor › Private AI › Generate Unit Tests

Context-aware prompts

File name, file type and language are added to each request automatically, so the model knows what it's looking at.

No prompt engineering required

Insert at caret or copy

Every code block in a response has one-click actions to drop it at your cursor or copy it to the clipboard.

From answer to editor in one click
How it works

Up and running in three steps

No sign-up, API key or credit card. If Ollama is already running, you can be chatting in under a minute.

  1. Run a model locally with Ollama

    Install Ollama and pull a coding model. Private AI is tuned for Qwen2.5-Coder, and any chat model Ollama serves will work.

  2. Install Private AI in IntelliJ IDEA

    Add the plugin under Settings › Plugins. The defaults point at http://localhost:11434 and qwen2.5-coder:latest, so most setups need no configuration.

  3. Ask, explain, refactor, test

    Open the Private AI tool window to chat, or right-click any selection and choose an action from the Private AI menu.

Terminal 
# 1. Pull a coding model (one-time download)
$ ollama pull qwen2.5-coder

# 2. Make sure Ollama is serving
$ ollama list
NAME                    SIZE
qwen2.5-coder:latest    4.7 GB

# 3. Optional: unplug the network. Private AI keeps working.
Fine-tune in Settings › Tools › Private AI: Ollama base URL, model name, request timeout and temperature.
Who it's for

For teams who can't paste code into a chatbot

If your security review, client contract or regulator says code stays in-house, Private AI lets you use AI assistance without asking for an exception.

Regulated industries

Finance, healthcare, government and defence teams can use AI assistance without new data-processing agreements or a third-party risk review.

Agencies & consultants

Client code under NDA stays on your workstation. You never have to explain to a customer that their source passed through another company's API.

Proprietary & pre-release IP

Unreleased products, trading algorithms and core IP get AI help without ever being uploaded anywhere.

Offline & restricted environments

Air-gapped networks, secure labs, flights and patchy connections. A local model behaves the same whether or not you're online.

Pricing

Try everything free for 30 days

The trial starts when you install the plugin and includes every feature. No account and no card.

Free trial

30 days

Everything in Private AI, from the first launch.

  • Chat tool window and all editor actions
  • Any Ollama model, local or on your network
  • No sign-up, no payment details
  • A reminder in the IDE before the trial ends
Start the trial

Private AI license

Pro

Keep using Private AI after the trial with a license key.

  • Everything in the trial, without a time limit
  • License key checked offline, with no phone-home
  • Activate in Settings › Tools › Private AI
  • No per-token fees, ever: you run the model
Get started
FAQ

Questions, answered

Does any of my code leave my machine?

No. Private AI sends requests only to the Ollama base URL set in its settings, which defaults to http://localhost:11434 on your own computer. It has no telemetry, analytics or other outbound connections.

If you change the base URL to an Ollama server elsewhere, for example a shared GPU box on your office network, requests go to that server and nowhere else.

Does it work without an internet connection?

Yes. You need a connection once, to download Ollama and a model. After that, chat, editor actions and license checks all work fully offline.

Which models can I use?

Any chat model Ollama can serve. The defaults and prompts are tuned for Qwen2.5-Coder, which gives strong coding results at sizes that fit on a developer workstation. You can switch models at any time in Settings › Tools › Private AI.

What hardware do I need?

Whatever your chosen model needs in Ollama. Smaller coding models (around 7B parameters) run on many modern laptops. Larger models give better answers but need more memory, ideally a GPU. Ollama's documentation lists requirements per model, and you can always point Private AI at a more powerful machine on your network.

Which IDEs are supported?

Private AI is built for IntelliJ IDEA 2025.3 and later.

How does licensing work?

Every install starts a 30-day trial with all features. When it ends, enter a license key in Settings › Tools › Private AI to keep using the plugin. The key is validated locally, so licensing never needs a network connection and never reports anything about you or your code.

Is a local model as good as a cloud model?

For everyday work like explaining code, suggesting refactors and drafting tests, current open coding models such as Qwen2.5-Coder do very well. The largest cloud models can still be ahead on the hardest tasks. Private AI is for teams who would rather have a capable assistant they fully control than send code off-site.

Use AI on code you can't share.

Install Private AI, connect it to your local model, and start your 30-day trial. It needs no account, no cloud and no compromises.

Get started Read our privacy policy