Give MCP agents structured graph memory with RushDB
Connect RushDB's MCP server so agents can store, search, update, and traverse persistent structured memory without hand-building a separate vector and graph stack.
npx skills add agentskillexchange/skills --skill give-mcp-agents-structured-graph-memory-with-rushdb
Use RushDB when an MCP-compatible agent needs persistent structured memory that can be searched semantically and traversed as relationships. The operator connects the RushDB MCP server, supplies a scoped RushDB API key or self-hosted API endpoint, and lets the agent create records, find records, inspect labels and properties, attach relationships, and recall prior decisions through the documented MCP tools. Invoke this instead of using RushDB as a normal database when the job is to give an agent a reviewable memory layer with discovery-first querying, not to build a general application backend. The scope boundary is MCP-based agent memory and schema-aware recall; it is not a generic database listing, a broad SDK reference, or a replacement for application data modeling.
What this skill actually does
Inputs and prerequisites: RushDB account or self-hosted RushDB API, RushDB API key, Node.js/npx, and an MCP-compatible client such as ChatGPT, Claude.ai, Claude Desktop, Cursor, Windsurf, VS Code, or Gemini CLI.
Setup notes: For local stdio MCP clients, add a rushdb MCP server entry that runs npx @rushdb/mcp-server and set RUSHDB_API_KEY in the server environment. For hosted MCP clients that support remote connectors, use the documented RushDB remote MCP endpoint and complete OAuth authorization. Use RUSHDB_API_URL only when pointing the server at a self-hosted or staging RushDB API.
Source and verification boundary: use https://github.com/rush-db/rushdb/tree/main/packages/mcp-server as the canonical reference before running the workflow; keep commands, API calls, CLI usage, and generated outputs reviewable against that upstream source.
Framework fit: publish this as a MCP workflow only when the operator can invoke the documented toolchain directly, rather than treating the upstream project as a generic product listing.