Shared memory MCP server for AI coding agents. Works identically with Claude Code and Cursor β same package, same config
threadctx is an MCP server that gives Cursor (and Claude Code) a memory: two tools, memory_write and memory_query, that your agent calls to record and recall project-specific learnings. Free and local-only by default β no signup, no network calls beyond your own LLM provider. Turn on the Team plan when you want that memory shared across everyone on the repo, with real semantic search instead of local keyword matching.
File location: .claude/mcp.json (project root) or ~/.claude/mcp.json (global)
{
"threadctxMcp": {
"command": "npx",
"args": [
"-y",
"threadctx-mcp"
]
}
}If you maintain threadctx-mcp, add this badge to your README to show it's verified on CuratedMCP:
[](https://www.curatedmcp.com/marketplace/threadctx-mcp)