Connect SuperLocalMemory to Cursor via MCP. Project decisions, code conventions, and debugging history persist across sessions in a local SQLite store — verify the integration contract in the installed Cursor release.
Package installation is separate from MCP configuration. Review the proposed Cursor MCP settings change before applying, then run a remember → recall round trip to confirm the integration.
Use pipx or uv with Python 3.11–3.14. Mode A does not require a model-provider key for its core path.
pipx install superlocalmemory slm setup
Run the connect command to review and apply the Cursor MCP configuration. Review the proposed settings change before confirming.
slm connect cursor
Or add manually in Cursor: go to Settings → Features → MCP Servers and add the server:
{ "mcpServers": { "superlocalmemory": { "command": "slm", "args": ["mcp"] } } }
Restart Cursor. In a Composer session, ask the agent to call get_status via the superlocalmemory MCP server. Then store a fact and recall it to confirm the round trip.
slm status # Expected: mode: A · provider: none · fact_count: 0 # Then try a round trip: slm remember "Frontend uses Next.js App Router with server components" slm recall "frontend architecture"
Memory state persists across Cursor sessions in a local SQLite store. Each capability below reflects documented SLM behavior — verify the integration contract for the installed release.
Cursor can store and selectively recall project decisions, folder conventions, and architectural choices through configured SLM tools. Retrieval is not complete-project recall.
Documented clients can target the same configured SLM profile. Each client handoff still needs identity, scope, lifecycle, and restart verification.
Core memory is SQLite-backed in the configured data root. Cursor, model providers, connectors, backups, proxies, and downloads retain separate network behavior and require separate assessment.
Local retrieval traces report candidate sources, fusion, reranker status, and measured latency. Performance depends on corpus, hardware, warmup state, and configuration.
Full source code is public and auditable. Verify the installed release, optional dependencies, configured providers, connectors, and the Cursor client behavior.
Optional behavioral signals can be surfaced as bounded context when configured. SLM does not by itself demonstrate that Cursor coding suggestions improve.
slm setup, and then run slm connect cursor to review the proposed MCP settings change before applying it. Then verify remember, recall, update, forget, reconnect, and uninstall behavior in your installed Cursor release.Install, connect, then verify the memory round trip in Cursor before treating it as active.