You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Context
Cursor already keeps its chat history in the local SQLite database state.vscdb, which contains tables such as messages and conversations. In cursor10x-mcp, however, every turn is sent to the MCP server and duplicated in a Turso-hosted SQLite database. As a result, the model spends extra tokens on *_storeMessage calls, and the data ends up in a second database that is not integrated with Cursor’s built-in history.
Question
Could you please explain why you decided not to leverage Cursor’s existing SQLite chat database for message storage, opting instead to create a separate Turso database and have the AI agent rewrite every message? Were there technical or architectural constraints that prevented working directly with state.vscdb (for example, via Cursor’s internal API or a local SQLite connection)? I’d love to understand the motivation and any benefits you see in this approach.