Agentset MCP Server

Content

Agentset MCP Server is the official @agentset/mcp package for connecting Claude to an Agentset knowledge base. It runs as a stdio MCP server and exposes a knowledge-base-retrieve tool that searches a configured namespace, returns matching chunks, and optionally reranks results.

Use it when an Agentset namespace already contains the documents Claude should consult, such as product documentation, support articles, policies, or internal technical knowledge. The server requires an Agentset API key and namespace ID, and it can optionally pass a tenant ID for multi-tenant data segregation.

Source Review

• https://github.com/agentset-ai/mcp-server
• https://docs.agentset.ai/production/mcp-server.md
• https://registry.npmjs.org/@agentset/mcp
• https://raw.githubusercontent.com/agentset-ai/mcp-server/main/README.md
• https://raw.githubusercontent.com/agentset-ai/mcp-server/main/LICENSE
• https://raw.githubusercontent.com/agentset-ai/mcp-server/main/package.json
• https://raw.githubusercontent.com/agentset-ai/mcp-server/main/src/index.ts
• https://raw.githubusercontent.com/agentset-ai/mcp-server/main/src/utils.ts
• https://raw.githubusercontent.com/agentset-ai/agentset/main/README.md
• https://docs.agentset.ai/search-and-retrieval/search.md

These sources were reviewed on 2026-06-06. Prefer the live MCP docs, NPM package metadata, MCP server source, option parser, Agentset platform README, and search documentation for current package versions and retrieval behavior.

Features

• Run the official @agentset/mcp package with npx, pnpm dlx, yarn dlx, or bunx.
• Search an Agentset namespace through the knowledge-base-retrieve MCP tool.
• Pass namespace ID through --ns or AGENTSET_NAMESPACE_ID.
• Pass a tenant ID with -t for multi-tenant data segregation.
• Override the default tool description with -d.
• Control result count with topK, capped by the server schema.
• Enable or disable reranking per retrieval call.
• Use Agentset's hosted API or a self-hosted Agentset deployment depending on your environment.

Installation

Run the MCP package with an Agentset API key and namespace:

AGENTSET_API_KEY=your-api-key npx @agentset/mcp --ns your-namespace-id

Add the server to Claude or another MCP client:

{
  "mcpServers": {
    "agentset": {
      "command": "npx",
      "args": ["-y", "@agentset/mcp@latest"],
      "env": {
        "AGENTSET_API_KEY": "agentset_xxx",
        "AGENTSET_NAMESPACE_ID": "ns_xxx"
      }
    }
  }
}

For multi-tenant applications, add the documented tenant option only after confirming the namespace and tenant mapping.

Use Cases

• Let Claude answer questions from an Agentset product documentation namespace.
• Retrieve relevant support articles with citations before drafting a response.
• Give an internal assistant access to policies or procedures without exposing unrelated namespaces.
• Use tenant scoping for customer-specific knowledge bases.
• Override the tool description so Claude knows when to use Agentset for a specific workflow.

Safety and Privacy

Agentset MCP is a retrieval bridge into a knowledge base, so access boundaries matter. Use least-privilege API keys, connect only the namespace Claude should query, and check tenant IDs carefully in multi-tenant applications. If a shared agent uses the server, document which namespace and tenant scope it can access.

Treat retrieval queries, chunks, citations, namespace IDs, tenant IDs, and API keys as sensitive. Retrieved content can include private documentation, support history, policies, product details, or customer data. Keep keys in environment variables or a secret manager, rotate them when shared configs change, and avoid copying private retrieval output into public issues, screenshots, or commits.

Duplicate Check

No agentset-ai/mcp-server, agentset-ai/agentset, @agentset/mcp, Agentset MCP Server, or matching source URL entry was found in content/mcp or README.md.