dropdat MCP server for Cursor

Give Cursor long-term, cross-session memory. Recall past conversations, save the current one, drop context into any AI client — all from inside Cursor. Cursor's native MCP support means a single config block gives the editor memory across every session and project.

Install in Cursor View on GitHub

Hybrid semantic recall

Vector cosine + BM25 fused with RRF. Asks in any phrasing, finds capsules that used different words.

Five MCP tools

dropdat_recall, dropdat_read, dropdat_list, dropdat_capsule, dropdat_autocapsule — all the surface an agent needs.

Cross-AI capsules

The same capsule library is reachable from Claude Code, Cursor, Cline, Claude Desktop, ChatGPT, Claude, Gemini.

Owned data

Capsules live in your dropdat library. Postgres + pgvector. Export anytime.

Install in Cursor

Two steps. Under a minute. No fork, no proxy.

01
Get a dropdat API key
Sign in at capsule.dropdat.app/api-keys and create a token (shape dk_live_…).

Install @dropdat/mcp

# zero-install
npx -y @dropdat/mcp

# or pin globally
npm install -g @dropdat/mcp

Drop config into ~/.cursor/mcp.json (or Settings → MCP → Add server)

{
  "mcpServers": {
    "dropdat": {
      "command": "npx",
      "args": ["-y", "@dropdat/mcp"],
      "env": {
        "DROPDAT_API_KEY": "dk_live_xxx",
        "DROPDAT_API_BASE": "https://dropdat.app"
      }
    }
  }
}

04
Verify
Open Cursor settings → MCP. The dropdat server lists five tools. Ask the chat 'Search dropdat for last week's API design notes.'

What you can ask Cursor now

Once wired in, any of these prompts route through dropdat tools.

“Find the dropdat capsule where we sketched the payment webhook flow.”
“Save the last 50 messages of this Cursor session as a capsule.”
“What did the ChatGPT conversation about Postgres indexes recommend? Check dropdat.”
“Continue from capsule xyz789 — load its full contents into context.”

Wire Cursor into your dropdat library.

Create a free account Compare all MCP clients