Summary
A single retrieval tool, and the customisable description is the part that makes it work.
A model chooses tools by reading them, and a generic "look up information in the knowledge base" tells it nothing about whether your HR policy lives there. Replace that text with what the namespace actually covers and the retrieval starts firing at the right moments instead of being ignored.
What it is
An MCP server for Agentset, an open-source retrieval-augmented generation platform. A single tool queries the namespace of documents you indexed and returns the passages that match.
What you get
- knowledge-base-retrieve — search the namespace you configured and read back the relevant passages
- Whatever you indexed is in scope: policies, procedures, product specifications, internal guidelines
- The tool's description can be replaced with your own wording, so the model knows what this particular knowledge base holds
- A tenant identifier can be supplied, so one deployment serves separated data
Requirements
An Agentset account, an API key, and the identifier of the namespace to query. Runs on Node.
Setup effort
One command plus a key — AGENTSET_API_KEY=your-api-key npx @agentset/mcp --ns your-namespace-id, then supply credentials