Shared memory for Cursor

Why Cursor teams add threadctx

• Team memory shows up in the agent panel. Cursor renders MCP tool calls and results directly in its Agent UI. When memory_queryreturns a hit, the recalled context and its · via threadctx attribution appear right where you're already reading the agent's output.
• It bridges to Claude Code, not just other Cursor users. threadctx is one MCP server behind both clients. A teammate running Claude Code in the terminal writes to the same memory a Cursor user reads from — cross-tool sharing that neither client does natively.
• Setup is the same JSON block Claude Code uses. No separate integration to build or maintain — MCP is a portable, open protocol, so adding threadctx to Cursor is a five-minute copy-paste.

Add it to Cursor

Add this block to .cursor/mcp.json in your project root, or via Cursor Settings → Tools & MCP. For local mode, drop the env block — no key needed.

{
  "mcpServers": {
    "threadctx": {
      "command": "npx",
      "args": ["-y", "threadctx-mcp"],
      "env": {
        "THREADCTX_MODE": "cloud",
        "THREADCTX_API_KEY": "tctx_..."
      }
    }
  }
}

v0 ships the plain-text attribution footer in Cursor's panel, matching Claude Code exactly. A clickable attribution chip linking to a public per-repo memory page is on the roadmap once there's a hosted page worth linking to.