Skip to content

Integration: JetBrains AI

PathStatusNotes
MCP server✅ SupportedJetBrains AI MCP client (since IDE version 2024.3)
HarnessNot supported; JetBrains AI routes through JetBrains’ cloud
ExtensionPlugin-based; use MCP

Confidence: High. JetBrains added MCP support in their November 2025 update (2024.3 release line). Tested on IntelliJ IDEA Ultimate, PyCharm Professional, and WebStorm.

Caveat: JetBrains AI routes generation through JetBrains’ own cloud infrastructure. MemStrata compresses the context you inject via MCP — it does not intercept the internal JetBrains routing for AI completion. For full context control, the MCP path is what’s available.


1. Start the MemStrata MCP server

Terminal window
memstrata serve --port 8080

2. Register in JetBrains IDE

Go to Settings / Preferences → Tools → AI Assistant → MCP Servers → Add:

FieldValue
NameMemStrata
URLhttp://localhost:8080/mcp
Enabled

Click Apply and Test Connection. You should see “Connected” with a list of available tools.

3. Use MemStrata context in AI Chat

In the AI Assistant panel, use the @memstrata mention to inject context:

@memstrata Explain the relationship between UserRepository and AuthService in this codebase.

MemStrata returns a compressed, graph-aware context slice. The AI Assistant receives this as a tool result and uses it in its response.


IDEMinimum version
IntelliJ IDEA Ultimate2024.3
PyCharm Professional2024.3
WebStorm2024.3
GoLand2024.3
Rider2024.3
PhpStorm2024.3

Community editions do not include AI Assistant and therefore do not support MCP.


SignalCaptured?Notes
AI Chat turns with @memstrata✅ YesFull context compression via MCP
Inline AI completions❌ NoInternal JetBrains routing; not MCP-accessible
AI Chat without @memstrata❌ NoJetBrains only calls MCP tools when explicitly invoked
Commit message generation❌ NoSeparate AI feature, not MCP-accessible

Note: Unlike Claude Code or Cline, JetBrains AI requires explicit @memstrata invocation. It does not automatically call MemStrata for every AI turn. This is a limitation of how JetBrains implements MCP — they treat it as an on-demand tool, not an ambient context provider.


“Test Connection” fails
Verify the harness is running: curl http://localhost:8080/health. If it returns {"status":"ok"}, the issue is JetBrains’ network settings — check that localhost connections aren’t blocked.

@memstrata not recognized
The MCP server must be registered and show as “Connected” in Settings. Restart the IDE after adding the server.

Context is stale
Run memstrata index to refresh the codebase graph. The JetBrains plugin does not trigger re-indexing automatically.