Guide
Cursor for JetBrains & IntelliJ: ACP Setup (2026)
Cursor runs inside JetBrains IDEs through the Agent Client Protocol (ACP). Open the AI chat panel in IntelliJ or PyCharm, choose Install from the ACP register, type Cursor and install. You need a JetBrains license and a Cursor account, not JetBrains AI premium. It is powered by the Cursor CLI, so the harness and models match the desktop app.
On this page
- What is the Cursor JetBrains integration?
- How do I install Cursor in IntelliJ?
- Do I need to pay for JetBrains AI to use it?
- Which modes and models are available in JetBrains?
- How do I set up MCP servers in JetBrains?
- What does JetBrains lack vs the Cursor desktop app?
- How do rules, skills and the CLI work here?
- When should I use JetBrains over Cursor desktop?
What is the Cursor JetBrains integration?
It is Cursor's coding agent running inside a JetBrains IDE through the Agent Client ProtocolAn open standard that lets any IDE host any coding agent; Cursor uses it to run inside JetBrains/IntelliJ and Android Studio via the Cursor CLI bridge. (ACPAgent Client Protocol. An open standard that lets any IDE host any coding agent; Cursor uses it to run inside JetBrains/IntelliJ and Android Studio.). The same integration works across the JetBrains lineup, so IntelliJ, PyCharm and Android Studio users get the agent without leaving their editor. Cursor's harness and models are identical to the desktop app; what changes is the surface it runs in.
The integration is an ACPAgent Client Protocol. An open standard that lets any IDE host any coding agent; Cursor uses it to run inside JetBrains/IntelliJ and Android Studio. extension, not a native app, so the desktop flow is smoother. The engine underneath is the same.
The underlying harness and model that's being used within the ACPAgent Client Protocol. An open standard that lets any IDE host any coding agent; Cursor uses it to run inside JetBrains/IntelliJ and Android Studio. is exactly the same as the IDE that we use with Cursor.
One mental model carries most of this guide: the JetBrains integration is powered by the Cursor CLICursor's command line: the full agent, all modes and models, in the terminal and pipeable into scripts and CI.. MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. configuration, slash commandsSaved multi-step prompts you trigger from the / menu, such as run tests, fix CI or open a pull request., the rules / skills / sub-agents primitives and model selection all behave the way they do in the CLI. If you know the CLI, you already know how the JetBrains surface is wired.
How do I install Cursor in IntelliJ?
Install it from inside the IDE's AI chat panel in under a minute. Open the AI chat (right sidebar), use Install from the ACPAgent Client Protocol. An open standard that lets any IDE host any coding agent; Cursor uses it to run inside JetBrains/IntelliJ and Android Studio. register, find Cursor and install it. Then say hello to confirm the agent answers.
- 1Install the Cursor CLICursor's command line: the full agent, all modes and models, in the terminal and pipeable into scripts and CI. from the Cursor downloads page (the ACPAgent Client Protocol. An open standard that lets any IDE host any coding agent; Cursor uses it to run inside JetBrains/IntelliJ and Android Studio. integration runs on it).
- 2In IntelliJ, open the AI chat panel from the right-hand sidebar.
- 3Choose Install from the ACPAgent Client Protocol. An open standard that lets any IDE host any coding agent; Cursor uses it to run inside JetBrains/IntelliJ and Android Studio. register, type Cursor and click install.
- 4Send a quick message like "hello Cursor" to confirm the agent responds.
- 5Pick a mode (Ask, Plan or Agent) and a model from the right-hand model picker, then start working.
Interactive diagram. Tab through its regions; each focused region shows its detail in the panel below.
The same five steps in IntelliJ, PyCharm or Android Studio. MCP is a separate, manual step covered below.
Do I need to pay for JetBrains AI to use it?
No. You need a JetBrains license for the IDE and a Cursor account, that is all. JetBrains sells its own AI premium tier, and you do not need it for this integration. The Cursor agent is billed and authenticated through Cursor, separately from any JetBrains AI subscription.
A Cursor team member fielded this exact question during the workshop.
As long as you have an IntelliJ license to be able to use the IDE, you can use this agent. You need a Cursor account as well. But you don't need to go for that extra piece there.
- You need
- A JetBrains IDE license (any tier)
- You do not need
- JetBrains AI premium / AI Assistant
- You need
- A Cursor account
- You do not need
- The Cursor desktop app
- You need
- The Cursor CLICursor's command line: the full agent, all modes and models, in the terminal and pipeable into scripts and CI. installed
- You do not need
- A separate per-IDE Cursor purchase
| You need | You do not need |
|---|---|
| A JetBrains IDE license (any tier) | JetBrains AI premium / AI Assistant |
| A Cursor account | The Cursor desktop app |
| The Cursor CLICursor's command line: the full agent, all modes and models, in the terminal and pipeable into scripts and CI. installed | A separate per-IDE Cursor purchase |
The integration rides on your existing JetBrains license plus a Cursor account.
Which modes and models are available in JetBrains?
Ask, Plan and Agent modes are all there, and you can switch modes and models mid-conversation in the same chat. Debug modeA mode that diagnoses a failure: it reproduces the issue, adds instrumentation and watches the logs, rather than reviewing a pull request., which exists in the Cursor desktop app, is not in the ACPAgent Client Protocol. An open standard that lets any IDE host any coding agent; Cursor uses it to run inside JetBrains/IntelliJ and Android Studio. integration yet. Cursor has said it is working with JetBrains to add it.
- Mode
- Ask
- In JetBrains ACP?
- Yes
- Use it for
- Read-only questions: project overview, how to run, where code lives
- Mode
- Plan
- In JetBrains ACP?
- Yes
- Use it for
- A reviewable plan before edits, best with a high-reasoning model
- Mode
- Agent
- In JetBrains ACP?
- Yes
- Use it for
- Multi-step edits and command execution
- Mode
- Debug
- In JetBrains ACP?
- Not yet
- Use it for
- Available in the desktop app; pending for ACPAgent Client Protocol. An open standard that lets any IDE host any coding agent; Cursor uses it to run inside JetBrains/IntelliJ and Android Studio.
| Mode | In JetBrains ACP? | Use it for |
|---|---|---|
| Ask | Yes | Read-only questions: project overview, how to run, where code lives |
| Plan | Yes | A reviewable plan before edits, best with a high-reasoning model |
| Agent | Yes | Multi-step edits and command execution |
| Debug | Not yet | Available in the desktop app; pending for ACPAgent Client Protocol. An open standard that lets any IDE host any coding agent; Cursor uses it to run inside JetBrains/IntelliJ and Android Studio. |
Mode availability as described in the 2026 JetBrains workshop. Check current docs before relying on parity.
On models, the same picker the CLI exposes is available here. Match the model to the job: a high-reasoning model in Plan modeA mode that makes no edits: it researches the codebase and produces an editable plan you review before any code changes., a fast in-house ComposerCursor's own fast coding model, tuned for the editor and priced well below frontier models; the recommended day-to-day model for executing a plan. model for quick edits. The model-selection logic is identical to the rest of Cursor.
Interactive widget. Tab through its controls; the result updates in the panel below as you change them.
Plan with a thinker, execute with a fast model.
High-reasoning models for planning and audits; fast executors for edits in a handful of files. The same routing applies in JetBrains.
If you are a heavy Cursor desktop user, the missing Debug modeA mode that diagnoses a failure: it reproduces the issue, adds instrumentation and watches the logs, rather than reviewing a pull request. is the first thing you will notice in JetBrains. Plan, Ask and Agent cover most workflows; reach for the desktop app or CLI when you specifically want Debug mode.
How do I set up MCP servers in JetBrains?
MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. setup in JetBrains is manual, unlike the one-click marketplace in Cursor and VS Code. Two steps are required, and the second one is easy to miss. Write your servers into an mcp.json at the user root, then enable MCP in IntelliJ's settings, which is off by default as a security measure.
- 1Create a .cursor/mcp.json at your user root level and add your MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. server config there.
- 2In IntelliJ, open Settings and search "MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. server".
- 3Check the box to enable MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs.. IntelliJ disables it by default as a security feature.
- 4For auth, pass tokens directly in mcp.json, or use the OAuth login flow for MCPs that need it.
The presenter lost half an hour to the hidden enable checkbox before MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. would connect at all.
The second gotcha which took me about 30 minutes the other day to find out is you have to go to IntelliJ settings, type in MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. server, you have to check mark this. It's a security feature within IntelliJ that doesn't enable MCP by default.
Because the integration is powered by the Cursor CLICursor's command line: the full agent, all modes and models, in the terminal and pipeable into scripts and CI., MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. configuration mirrors the CLI almost exactly. For OAuth-based servers you can authenticate the same way you would in the CLI; the most reliable first pass is still an API key set directly in mcp.json. If MCP tools do not appear, check the enable checkbox before anything else.
What does JetBrains lack vs the Cursor desktop app?
The JetBrains integration is an ACPAgent Client Protocol. An open standard that lets any IDE host any coding agent; Cursor uses it to run inside JetBrains/IntelliJ and Android Studio. extension, while the Cursor desktop app is built AI-native from the ground up, so the desktop flow is smoother and a few features land there first. The agent's intelligence is the same; the surface is thinner. Be honest about that when you choose where to work.
- Area
- Underlying harness + models
- Cursor desktop
- Native
- JetBrains ACP
- Identical engine via the CLI
- Area
- Ask / Plan / Agent modes
- Cursor desktop
- Yes
- JetBrains ACP
- Yes
- Area
- Debug modeA mode that diagnoses a failure: it reproduces the issue, adds instrumentation and watches the logs, rather than reviewing a pull request.
- Cursor desktop
- Yes
- JetBrains ACP
- Not yet
- Area
- MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. install
- Cursor desktop
- One-click marketplace
- JetBrains ACP
- Manual mcp.json + enable checkbox
- Area
- Overall feel
- Cursor desktop
- AI-native, smoothest
- JetBrains ACP
- Extension; a step behind on polish
| Area | Cursor desktop | JetBrains ACP |
|---|---|---|
| Underlying harness + models | Native | Identical engine via the CLI |
| Ask / Plan / Agent modes | Yes | Yes |
| Debug modeA mode that diagnoses a failure: it reproduces the issue, adds instrumentation and watches the logs, rather than reviewing a pull request. | Yes | Not yet |
| MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. install | One-click marketplace | Manual mcp.json + enable checkbox |
| Overall feel | AI-native, smoothest | Extension; a step behind on polish |
Parity as described in the 2026 workshop. The gaps are surface-level, not model-level.
Codebase understanding carries over. Cursor indexes your repo with semantic parsing into a vector database and combines that with grep, which is why an Ask-mode "give me an overview of this project and how to run it" returns fast even in an unfamiliar codebase. That indexing works the same inside JetBrains.
How do rules, skills and the CLI work here?
The same primitives you use across Cursor apply in JetBrains, because the integration runs on the CLI. Rules are the standards the agent must follow. Skills are an on-demand playbook that both you and the agent can invoke. Sub-agents are specialists. You author and call them the same way you would in the CLI or desktop.
Interactive widget. Tab through its controls; the result updates in the panel below as you change them.
Reach for the lightest primitive that does the job; spend context deliberately.
What each primitive is, who invokes it and how much control you keep. Available in JetBrains via the CLI engine.
Because the integration shares the CLI engine, you can keep the IDE for review and use the terminal for fan-out. Open several terminal tabs in the same repo, start an agent in each with a different model, and give them the same task to compare implementations. The IDE handles visual review; the CLI handles parallel runs.
Keep rules short and scoped. Cursor's own open-source rules run around eight lines, and a rule set over a hundred lines is a signal to trim it or split it across file-scoped rules. Context is a scarce resource, so a rule that only fires on the files it applies to beats one that runs on every request.
When should I use JetBrains over Cursor desktop?
Use the JetBrains integration when your team already lives in IntelliJ, PyCharm or Android Studio and switching editors is a real cost. The agent meets them where they are, with the same models and harness. Reach for the desktop app or CLI when you need Debug modeA mode that diagnoses a failure: it reproduces the issue, adds instrumentation and watches the logs, rather than reviewing a pull request., the smoothest AI-native flow or one-click MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. installs.
After install, run a quick Ask-mode project overview to confirm indexing works, then add one MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. server and verify it appears (remember the enable checkbox). Setup commands and the ACPAgent Client Protocol. An open standard that lets any IDE host any coding agent; Cursor uses it to run inside JetBrains/IntelliJ and Android Studio. register version change over time, so check the current Cursor docs before wiring it into a team workflow.
Frequently asked questions
Does Cursor work in IntelliJ and PyCharm?
Yes. Through the Agent Client Protocol (ACP), Cursor's agent runs inside JetBrains IDEs including IntelliJ, PyCharm and Android Studio. Install it from the ACP register in the IDE's AI chat panel. The harness and models match the Cursor desktop app.
Do I need JetBrains AI premium to use Cursor in IntelliJ?
No. You need a JetBrains license for the IDE and a Cursor account. You do not need JetBrains' paid AI premium tier, and you do not need the Cursor desktop app. The Cursor CLI must be installed, since the integration is powered by it.
Why won't my MCP servers connect in IntelliJ?
Two things are required. Put your servers in a .cursor/mcp.json at the user root, then open IntelliJ Settings, search "MCP server" and tick the enable checkbox. IntelliJ disables MCP by default as a security feature, and that hidden checkbox is the most common reason MCP tools never appear.
Is Debug mode available in the JetBrains integration?
Not yet. The JetBrains ACP integration offers Ask, Plan and Agent modes, and you can switch modes and models mid-chat. Debug mode exists in the Cursor desktop app and is being worked on for ACP. Use the desktop app or CLI when you specifically need it.
How is the JetBrains integration related to the Cursor CLI?
It is powered by the Cursor CLI. MCP configuration, slash commands, the rules / skills / sub-agents primitives and model selection all behave like the CLI. If you know the CLI, the JetBrains surface follows the same patterns, including the manual mcp.json setup.
Is anything different from the Cursor desktop app?
The intelligence is identical, but JetBrains runs as an ACP extension while the desktop app is AI-native, so the desktop flow is smoother. Specific gaps: Debug mode is not in JetBrains yet, and MCP install is manual rather than one-click. Codebase indexing and the models are the same.
Sources & last verified
Cursor ships frequently. Facts verified against primary sources on June 25, 2026.