JetBrains AI Assistant not working: causes and workarounds

The query "jetbrains ai assistant not working" covers several different problems: the plugin will not activate, the chat stays silent, features vanish in one project, or a model from your API does not show up in the list. Below are the causes described in the JetBrains AI Assistant documentation (edition 2026.2, pages updated in October 2026) and two ways out: repair the built-in service, or connect your own models through an OpenAI-compatible API. We take menu names from the documentation; in your IDE version they may differ slightly, so check against the screen.

JetBrains AI Assistant not working: seven causes

1. Plugin or widget

The documentation describes two levels of disabling: per project (the "Disable for This Project" item in the JetBrains AI widget) and for the whole IDE via the plugin. If the icon is missing, check the list of installed plugins. Without the plugin, the widget and the AI Chat tab remain only as an invitation to install it and have no access to your code.

2. The .noai file

An empty .noai file in the project root fully disables AI Assistant features, even if the project is opened in another IDE. Such a file often arrives with a repository from a team where AI is forbidden. It does not affect third-party AI plugins.

3. The .aiignore file

If the failure is only on some files, look at .aiignore. The syntax is the same as .gitignore. AI Assistant also honors .cursorignore, .codeiumignore and .aiexclude in the project root, so the rule may come from another tool.

4. Network

The documentation lists the service base addresses that can be blocked at the network level: https://api.jetbrains.ai/ and https://api.app.prod.grazie.aws.intellij.net/. If a corporate firewall, proxy or DNS blocks them, the chat will not answer. Check whether they open from the same network where the IDE runs.

5. Activation

The documentation describes several activation options, including a JetBrains AI subscription and your own API keys for third-party providers; see the activation section for the current full list. If none is chosen, there is nothing to work with.

6. Organization management

If AI is configured centrally in your organization (JetBrains IDE Services or Central), an administrator can set the provider and forbid connecting your own models. Then the choice is limited or unavailable, and only the administrator can change it.

7. Own models without assignment

The most common trap for people who already connected a key. According to the documentation, models from Anthropic, Google and OpenAI are assigned to features automatically, but only if the model is on the feature's predefined list. Otherwise the feature is unavailable. Models from an OpenAI-compatible endpoint are assigned manually in the Models Assignment block, and until then they work only in AI Chat.

JetBrains AI in Russia

The availability of the built-in JetBrains AI service and the ways to pay for a subscription depend on region and JetBrains terms, so read the current terms on the JetBrains site: we do not comment on them. If the built-in service is unavailable to you, the working option does not need it at all: the IDE can work with third-party models using your own key. From Russia you can reach the models through the gateway and top up via SBP, MasterCard or crypto.

Connecting your own models through Zukko

Steps according to the AI Assistant documentation:

1. Open Settings | Tools | AI Assistant | Providers & API keys.
2. In the Third-party AI providers block, choose a provider compatible with the OpenAI API (check the exact item name in your IDE's list).
3. In the URL field enter https://api.zukko.pro/v1.
4. In the API Key field paste your fetch_... key from the Zukko dashboard.
5. If the model must call tools of MCP servers, enable Tool calling.
6. Click Test Connection, then Apply.
7. Open AI Chat: the provider's models appear in a separate section of the selector.
8. In the Models Assignment block assign models to Core features and Instant helpers.

If Test Connection returns an error, start with the key (401 means a wrong key, 402 means insufficient balance) and the address. Whether the URL field in your IDE version needs /v1, check with this test: for OpenAI-format clients Zukko uses a base with /v1. More about the format: chat/completions↗.

Which models to assign

• Core features (in-editor generation, commit messages, the default model in AI Chat): claude-sonnet-5-5 or gpt-6.1-sol. This is a guide by price and strength index (48 and 52); whether a model suits you, test on your own code.
• Instant helpers (chat titles, name suggestions, context collection): a cheap model, for example deepseek-v4.1-flash ($0.013 / $0.038 per 1M) or glm-5.3-flash ($0.04 / $0.04).

What it cannot replace

According to the documentation, inline completion and next edit suggestions do not use the models from this block. For completion you can separately connect an OpenAI-compatible model, but it must be built for code completion (with fill-in-the-middle support), and general-purpose chat models are usually not suitable. So chat and code generation through the gateway work, while fast "as you type" completion stays on JetBrains' built-in models. To have all features available, the documentation recommends using your own models together with a JetBrains AI subscription.

What it costs

Prices as of October 18, 2026. Say (illustrative volumes) you send 20 chat requests a day, each with 6,000 input and 800 output tokens (including attached files). Formula: 6000 × input price / 1M + 800 × output price / 1M.

For Claude through the OpenAI format the whole input is billed at the input price and no cache discount applies. AI Assistant has a message trimming threshold: attachments that take more than the set share of the context window get trimmed so that the request fits the model. It is not a cost control: manage spend with a key limit. Current prices: GET https://api.zukko.pro/v1/models.

See also the detailed IDE guide: JetBrains AI Assistant: connecting in IntelliJ and PyCharm↗.

FAQ

My key works but AI Chat shows no models. Why?

Check that you clicked Apply after Test Connection and opened the model selector: provider models appear in a separate section. If the section is missing, check whether an organization administrator restricted connecting your own models.

Do I need a JetBrains AI subscription to use my own models?

Not for some features, but the documentation directly recommends combining your own models with a subscription so all features work. Subscription terms are on the JetBrains site.

Why does completion not work with my model?

Completion is configured separately in the AI Completion section and needs a code completion model. A chat model from the Third-party providers block is not used for it.

How do I cap spending quickly while I set this up?

Create a separate key for the IDE with a daily dollar limit and a list of allowed models. If the limit is used up, the API returns 429 with the code key_daily_limit_reached instead of charging extra.

Create a key in the dashboard↗, give it a limit and connect it using the steps above.