MemData

MemData MCP Server provides persistent memory for AI agents, allowing long-term memory across conversations by storing notes, decisions, and context and retrieving them semantically later. It enables Claude, Cursor, and any MCP-compatible client to remember everything between sessions and across hosts.

How to use

Use the MemData MCP Server with your MCP client to ingest and retrieve memories. You can store notes, decisions, and context, then query for relevant context in future sessions. The server supports both human-powered usage with an API key and autonomous agents that pay per operation.

How to install

Prerequisites: Node.js and npm installed on your system.

# Install the MemData MCP package locally (no global install required)
npm init -y
npm install memdata-mcp

Run the MCP server client using the provided CLI within your project. The configuration shows how to connect via stdio, invoking npx to run memdata-mcp and providing the required environment variables.

{
  "mcpServers": {
    "memdata": {
      "command": "npx",
      "args": ["memdata-mcp"],
      "env": {
        "MEMDATA_API_KEY": "md_your_key",
        "X402_WALLET_KEY": "your_private_key_hex",
        "MEMDATA_API_URL": "https://memdata.ai"
      }
    }
  }
}

Additional configuration and usage notes

Choose the appropriate authentication method for your setup. You can use a subscription API key for humans or a wallet-based pay-per-use flow for autonomous agents. If you enable the wallet-based flow, payments and identity are tied to the wallet address used by the agent.

Environment variables you may configure include: MEMDATA_API_KEY for subscribers, X402_WALLET_KEY for autonomous agents, and MEMDATA_API_URL to point to the MemData service if you self-host or use a custom endpoint.

Security and data handling

The MCP client acts as a thin layer that calls the MemData API. It does not store data locally, and it does not collect analytics or telemetry. You can inspect the source code in the project to understand data flow and handling.

Usage example

After configuring the client, you can ingest and query memories through your AI workflow.

You: Remember that we chose PostgreSQL for the user service
AI: [calls memdata_ingest] → Stored in memory

... days later ...

You: What database are we using for users?
AI: [calls memdata_query] → "PostgreSQL for the user service" (73% match)

Available tools

memdata_ingest

Store text in long-term memory by providing content and identity for source tracing.

memdata_query

Search memory with natural language and retrieve relevant results.

memdata_list

List all stored memories with associated chunk counts.

memdata_delete

Delete a memory by its artifact ID.

memdata_status

Check API health and current storage usage.

memdata_session_start

Initialize a session and obtain agent identity and recent activity.

memdata_set_identity

Set or update the agent's identity and identity summary.

memdata_session_end

Save session context and handoff for continuity.

memdata_query_timerange

Query memories within a specific date range.

memdata_relationships

Find related entities appearing together in memory.