JetBrains AI is the set of AI features, agents and subscriptions built into JetBrains IDEs, with AI Assistant as the in-IDE layer and Junie as JetBrains' own coding agent. The naming has moved quickly: AI Assistant, Junie, Junie CLI, Junie Lite, JetBrains Air and JetBrains Central all appear on JetBrains' own pages, and the credit-based plans changed during 2025 and 2026. This page sets out what each part is, what the AI Free, AI Pro, AI Ultimate and AI Enterprise tiers include, how AI credits work, which models and IDEs are supported, and how the product has developed. It is written for developers, engineering managers and the people who approve tools for a team. Every fact comes from JetBrains' own website, documentation, legal pages and blog. Last checked: 2026-11-15.
This is an independent guide from NHI Mgmt Group, based on JetBrains' public documentation as of November 2026. It is not sponsored or reviewed by JetBrains. Product features change, so confirm settings against JetBrains' current documentation.
Key takeaways
- Two products, one subscription. AI Assistant is the collection of AI features inside JetBrains IDEs; Junie is the coding agent that runs in the IDE, in a terminal (Junie CLI) and in CI through GitHub and GitLab. Both draw on the same JetBrains AI credits.
- Usage is measured in AI credits. JetBrains states that one credit equals 1 US dollar in local currency, that quota resets every 30 days, and that top-up credits stay valid for 12 months.
- Four tiers plus a trial. AI Free (3 credits per 30 days), AI Pro (10 credits), AI Ultimate (35 credits) and AI Enterprise for organisations through JetBrains IDE Services.
- Model choice is open. JetBrains routes requests to OpenAI, Google, AWS Bedrock, Anthropic and SpaceXAI models, hosts its own Mellum models and Qwen models, and supports your own API keys and local models.
- Junie left beta in June 2026 and is now part of JetBrains Air. The general availability release added plan mode, agentic debugging, remote control and code review, and rebuilt the agent on the Agent Client Protocol (ACP).
What is JetBrains AI Assistant and Junie?
JetBrains' documentation describes AI Assistant as "a collection of AI-powered features and coding agents integrated into JetBrains IDEs." It is delivered as a plugin in IntelliJ IDEA, PyCharm, WebStorm and the other JetBrains IDEs, and it groups its features into context-aware AI Chat, coding agents, in-editor assistance (code generation, inline completion and next edit suggestions), code insights (explanations, refactoring and issue detection) and routine automation (documentation, unit tests, commit messages and pull request summaries).
Junie is JetBrains' own coding agent. Its product page calls it "the coding agent that works with any model you choose." Where AI Assistant helps a developer write and understand code inside the editor, Junie takes a task, writes a plan, edits files across the project, runs tools and reports back. JetBrains' GA announcement states that the agent was rebuilt on ACP and is shared across the AI chat, the Junie tool window and Junie CLI, so the same agent appears in all three places.
Above both sits JetBrains Air. JetBrains' AI page describes Air as a system for building software with agents, available in JetBrains IDEs, on the web, in a CLI and on mobile, and states that Junie is "now part of JetBrains Air." Air runs Junie alongside other ACP agents, which is why the same subscription can pay for several agents.
AI Assistant can be activated in four ways, per the documentation: a JetBrains AI subscription, your own API keys, authorising an integrated agent with a provider account, or installing external coding agents. ACP agents installed from the ACP registry do not need a JetBrains AI subscription.
Key facts at a glance
As of 15 November 2026, from JetBrains' own pages.
| Field | Value | Source page |
|---|---|---|
| Vendor | JetBrains | JetBrains AI page |
| Product type | In-IDE AI features (AI Assistant) and coding agent (Junie) for IDE, terminal and CI | AI Assistant docs, Junie page |
| Junie status | Out of beta (GA announcement, June 2026); part of JetBrains Air | Junie blog, JetBrains AI page |
| Tiers | AI Trial, AI Free, AI Pro, AI Ultimate, AI Enterprise | Licensing help page |
| Usage unit | AI credits; 1 credit = 1 US dollar in local currency; 30-day quota | Licensing help page |
| Free tier | AI Free with 3 credits per 30 days; Junie Lite free model in every JetBrains account | Licensing help page, Junie page |
| Bundled licences | AI Pro included with All Products Pack, dotUltimate and DataSpell | Licensing help page |
| Model sources | OpenAI, Google, AWS (Bedrock), Anthropic, SpaceXAI; JetBrains-hosted Mellum, Mellum2 and Qwen | AI service providers page |
| Bring your own key | Yes; Junie page states "10+ models supported via BYOK" at provider rates with zero markup | Junie page |
| Local models | Yes; LiteLLM, LM Studio and Ollama named for Junie | Junie GA blog |
| Training on customer data | Not unless the customer agrees (AI Terms of Use, Section 5(c)) | JetBrains AI Terms of Use |
| Regional restriction | Service not available in China since 13 July 2026 | Licensing help page |
Plans and pricing
JetBrains sells AI as a separate subscription from the IDE licence, for individuals and for organisations. Each licence includes a quota of AI credits that cloud-model features consume. The credit allowances below are as of 15 November 2026.
| Tier | Credits per 30 days | Price shown by JetBrains | Who can buy |
|---|---|---|---|
| AI Trial | Trial allowance | Free | Individuals and organisations evaluating |
| AI Free | 3 | Free | Individuals and organisations |
| AI Pro | 10 | $8.33 per user per month on the Junie page; $10 on the licensing help page | Individuals and organisations |
| AI Ultimate | 35 | $25 per user per month on the Junie page; $30 on the licensing help page | Individuals and organisations |
| AI Enterprise | "On par with AI Ultimate quota, or higher" | Through JetBrains IDE Services | Organisations only |
Reading the prices: the Junie page shows an Annual and Monthly toggle beside its figures, and the licensing help page shows more than one price table as of 15 November 2026. Treat the credit allowances as the stable part and confirm the price for your billing period on JetBrains' pricing page before you budget.
How credits are used: JetBrains gives rough equivalents for one credit: about 10 AI Chat code generation requests, 40 in-editor code generations, 25 code explanations, 280 documentation requests or 140 commit messages. Token count drives the cost, so long chat threads, models marked as high cost and agent mode use credits faster. The quota resets 30 days after first use of the licence.
Top-ups: extra credits can be bought for AI Trial, AI Pro and AI Ultimate, are used only after the monthly quota is spent, and stay valid for 12 months. They are bought through the JetBrains AI widget in the IDE or the JetBrains Account. Top-ups are not offered for AI Free, and purchased credits are not available in China.
Bundles and restrictions: AI Pro comes with the All Products Pack, dotUltimate and DataSpell. AI Free is not available in Android Studio, in the Community editions of PyCharm and IntelliJ IDEA, or in PyCharm Unified used for free, and is not added automatically to IDE trial licences. Neither AI Free nor the bundled AI Pro can be activated when the IDE licence comes from a License Server.
Organisations on JetBrains Central: for organisations migrated to JetBrains Central, the documentation states that no separate AI subscription is needed; an admin grants AI access and sets default and per-user monthly credit limits from one shared pool. Migration allocations per licence per month are AI Ultimate 70, AI Pro 20, All Products Pack 20, dotUltimate 20 and AI Free 3.
Models available
JetBrains lists its AI service providers on a legal page and the current agent models on the Junie page. Providers are listed alphabetically below; the table is as of 15 November 2026.
| Source (alphabetical) | Models or role | Where stated |
|---|---|---|
| Alibaba Cloud (trainer) | Qwen models, run on JetBrains servers | AI service providers page |
| Anthropic | Claude models, directly or through AWS Bedrock; Junie page lists Claude Opus 5.5, Claude Sonnet 5.5 and Claude Fable 5.2 | Service providers page, Junie page |
| AWS | Amazon Bedrock models | AI service providers page |
| Gemini models; Junie page lists Gemini 4 Carbon, Gemini 4 Argon and Gemini 3.8 Flash | Service providers page, Junie page | |
| JetBrains | Mellum and Mellum2, trained by JetBrains and run on JetBrains servers; Junie Lite free model | Service providers page, Junie page |
| OpenAI | GPT models; Junie page lists GPT-6 ASTRA and GPT-6.1 SOL | Service providers page, Junie page |
| SpaceXAI | Grok models; Junie page lists Grok 4.7 | Service providers page, Junie page |
- Self-hosted models: for Mellum, Mellum2 and Qwen, JetBrains states that "their inputs and outputs are not shared with the parties that trained them." For Amazon Bedrock it states that inputs and outputs are not shared with the model providers.
- Bring your own key: the Junie GA post names Anthropic, OpenAI and Google among BYOK providers. Air's launch post states that your own keys are used first, with usage they do not cover falling back to the JetBrains subscription.
- Local models: Junie supports local runtimes through LiteLLM, LM Studio and Ollama; JetBrains states that prompts and code are then not shared externally.
- Plan on one model, build on another: the Junie page describes planning with one model and implementing with another to cut cost.
Model names follow each provider's release cycle and change often. Check the Junie page and the AI service providers page before naming a model in internal guidance.
Features and modes
| Feature | What it does (per JetBrains) | Product |
|---|---|---|
| AI Chat | Questions about code and project structure, answered with the current context | AI Assistant |
| Inline completion and next edit suggestions | In-editor completion and suggestions for the next change | AI Assistant |
| In-editor generation | Generate or update code from natural language prompts | AI Assistant |
| Code insights | Explanations, improvement ideas, refactoring help and issue detection | AI Assistant |
| Routine automation | Documentation, unit tests, commit messages and pull request summaries | AI Assistant |
| Advanced Plan Mode | Writes requirements, technical design and delivery stages to .junie/plans for review before coding | Junie |
| Live Prompting | Steer the agent while a task is running | Junie |
| Human in the Loop | Review and approve key actions, with a dynamic allowlist and user-confirmed execution | Junie |
| Agentic debugging | Starts or joins debug sessions, sets breakpoints (including in library and decompiled code) and inspects runtime state | Junie |
| Remote Control | Runs tasks asynchronously and lets you monitor them from any signed-in device | Junie |
| Code review | Reviews from GitHub Actions, GitLab (including on-premises) or the /review command | Junie |
| Guidelines, Agent Skills and commands | Persistent guidelines, Agent Skills packages and /commands shared across CLI and IDE through ACP | Junie |
| Database integration | Works with databases configured through DataGrip and the JetBrains Database plugin | Junie |
Benchmark claim: JetBrains' GA post reports that Junie placed first among coding agents on SWE-Rebench with 61.6 per cent resolved and 72.7 per cent pass@5. This is a vendor-reported figure from the announcement; check the benchmark's own site for the current table.
Supported IDEs and platforms
- JetBrains IDEs: the AI Assistant documentation covers CLion, DataGrip, DataSpell, GoLand, IntelliJ IDEA, PhpStorm, PyCharm, Rider, RubyMine, RustRover and WebStorm. It also mentions Android Studio and ReSharper, while stating that its documentation does not cover use in those products.
- IDE versions: AI Free and AI Ultimate are available in most IDEs from version 2025.1, with AI Ultimate working from 2024.2.1. Older IDE versions support AI Pro and AI Enterprise.
- Junie in the IDE: the GA post states support in "all JetBrains IDEs", through the AI chat and the Junie tool window, and through Air in IDEs for running Junie across several projects.
- Terminal: Junie CLI is "a command-line AI agent that helps you write code", installed on macOS, Linux and Windows with an install script.
- CI and code hosting: a Junie GitHub Action for issues and pull requests, and GitLab CI/CD for issues and merge requests.
- JetBrains Air: available in JetBrains IDEs, on the web, as a CLI and on mobile, per the JetBrains AI page. The March 2026 public preview launched on macOS first, with Windows and Linux announced as planned.
Enterprise and admin controls
| Control | Documented setting or feature | Where |
|---|---|---|
| Organisation licensing | Per-user AI licences; admins manage top-up credits in the JetBrains Account and set a monthly limit per subscription | JetBrains Account |
| Shared credit pool | Admin grants AI access and sets default and per-user monthly credit limits | JetBrains Central (migrated organisations) |
| Enterprise deployment | AI Enterprise (Pay As You Go), enabled for selected users in a profile, with a consumption dashboard widget | JetBrains IDE Services |
| Isolated infrastructure | AI Enterprise configured to operate in air-gapped environments | JetBrains IDE Services |
| Custom models | Custom OpenAI, Azure OpenAI and Gemini model configuration | JetBrains IDE Services |
| Agent and model permissions | Organisation-wide or team and individual permissions for models and agents; spending limits; BYOK from Amazon Bedrock, OpenAI, Anthropic or other providers; usage analytics | JetBrains Air (Air Teams in early access) |
| Project opt-out | Empty .noai file in the project root disables all AI Assistant features for that project in any JetBrains IDE | Repository |
| File exclusion | .aiignore with .gitignore syntax; .cursorignore, .codeiumignore and .aiexclude also honoured | Repository, Settings | Tools | AI Assistant | Project Settings |
| Network block | Block https://api.jetbrains.ai/ and https://api.app.prod.grazie.aws.intellij.net/ | Network |
The documentation notes that .noai does not affect third-party AI plugins and that files listed in .aiignore may occasionally still be processed, so neither replaces a policy on what code may reach an AI service. API keys used for BYOK, the Junie GitHub Action and GitLab CI jobs are non-human identities: give each an owner, a narrow scope and a rotation date. See AI Coding Tools and Secrets Sprawl and How to audit AI coding assistant tokens and config files.
Data handling and privacy
- Training: Section 5(c) of the JetBrains AI Terms of Use states that JetBrains will not use your data for training unless you agree to it, and that AI service providers are bound by substantially the same restrictions.
- Storage: the same section states that inputs, data and outputs may be stored and used for a limited period; the Terms do not give a number of days.
- Monitoring: inputs and outputs may be monitored by JetBrains or providers to prevent acceptable use violations.
- What counts as your data: source code, information derived from it, and usage-related information from the user's device sent with inputs (Section 2).
- Personal data: Section 13(b) states that providing personal data as input into JetBrains AI is prohibited, and personal data is otherwise processed under a Data Processing Addendum.
- Subprocessors: not every listed provider is involved in every request, per the AI service providers page.
- Assurance: the Junie page states that JetBrains tools adhere to security standards including SOC 2 certification.
For organisations, the practical decision is which providers are acceptable. Local models, JetBrains-hosted models or an air-gapped AI Enterprise deployment keep code inside a narrower boundary than cloud frontier models; BYOK moves the data relationship to your own provider contract.
Release history
Dated milestones from JetBrains' own pages. JetBrains' fetched pages do not state a first release date for AI Assistant, so the history starts with the dated items below.
| Date | Event | Details |
|---|---|---|
| 2024.2.1 IDE release | AI Ultimate support | Earliest IDE version stated for AI Ultimate |
| 2025.1 IDE release | AI Free and AI Ultimate in most IDEs | Free tier with 3 credits per 30 days available from this version |
| March 2026 | JetBrains Air public preview | macOS first; Codex, Claude Agent, Gemini CLI and Junie built in; ACP support |
| June 2026 | Junie leaves beta | Plan mode, agentic debugging, remote control, code review; rebuilt on ACP; database integration |
| 13 July 2026 | China | JetBrains AI service no longer available in China |
| As of November 2026 | Junie part of JetBrains Air | Air in IDEs, Air Teams (early access), Air Governance, Air Gateway and Air Context listed on the JetBrains AI page |
Limits and known constraints
- 30-day quota: included credits reset 30 days after first use; agent mode and high-cost models spend them fastest.
- AI Free scope: 3 credits, no top-ups, and not available in Android Studio or the Community editions of PyCharm and IntelliJ IDEA.
- License Server: AI Free and the bundled AI Pro cannot be activated when the IDE licence comes from a License Server.
- AI Enterprise top-ups: the documentation states that buying top-ups for AI Enterprise "will be implemented later".
- Territory: the service is restricted to allowed territories and has not been available in China since 13 July 2026.
- Exclusion files:
.aiignoreis best effort, and.noaicovers AI Assistant only, not third-party AI plugins. - Price display: record the date and billing period whenever you quote a JetBrains AI price.
Security guidance
This page covers what JetBrains AI and Junie are and cost. Hardening is covered separately in the JetBrains AI Assistant and Junie Security guide, which goes through identity and access, admin roles, API tokens and service accounts, logging and a hardening baseline. The short version for a buyer: choose the organisation licensing model (per-user, JetBrains Central or AI Enterprise) before rollout, set credit limits, decide which providers are allowed, commit .noai or .aiignore where code must stay out of AI services, and treat repository files that steer the agent, such as guidelines and .junie content, as untrusted input in pull requests. The wider discipline is in the AI IDE and Coding Agent Security Guide and the Coding Agent Configuration File Attacks guide.
Practitioner checklist
- Record the date, tier and billing period whenever you quote a JetBrains AI price, and recheck before renewal.
- Decide between per-user licences, JetBrains Central and AI Enterprise through IDE Services for organisational use.
- Set monthly credit limits per subscription or per user before enabling agent mode widely.
- Agree which model providers are allowed, and whether BYOK, local models or JetBrains-hosted models are required for sensitive code.
- Commit
.noaito repositories that must not reach AI services, and.aiignorefor sensitive paths, knowing both have limits. - Inventory API keys and CI tokens used by Junie CLI, the GitHub Action and GitLab jobs, with an owner and rotation date for each.
- Check IDE versions across the estate against the tier you buy (2025.1 for AI Free, 2024.2.1 for AI Ultimate).
- Review JetBrains' AI service providers page and AI Terms of Use when the provider list changes.
Standards and references
- JetBrains: AI Assistant licensing and subscriptions
- JetBrains: About AI Assistant
- JetBrains: Junie
- JetBrains Blog: Junie, the JetBrains AI Coding Agent, Leaves Beta
- JetBrains: JetBrains AI Terms of Use
- JetBrains: JetBrains AI service providers
- JetBrains: Disable AI Assistant
- JetBrains: AI Enterprise in IDE Services
Related NHI Mgmt Group resources: JetBrains AI Assistant and Junie Security · AI IDE and Coding Agent Security Guide · AI Coding Tools and Secrets Sprawl · AI Coding Assistant Buyer's Guide