MCP Server

Connect Claude, Cursor, and other AI agents to Agentset with the hosted MCP server, or run the lightweight local server for search-only access.

Agentset provides two MCP servers:

• Hosted MCP server — a remote server at https://api.agentset.ai/mcp that exposes the full Agentset platform. Agents can create namespaces, ingest data, search, chat, and manage your organization autonomously.
• Local MCP server — the @agentset/mcp npm package, a lightweight stdio server that searches a single namespace.

Hosted MCP server

The hosted server uses streamable HTTP and authenticates with your API key as a bearer token:

https://api.agentset.ai/mcp
Authorization: Bearer agentset_xxx

To scope requests to a single tenant in multi-tenant setups, add an optional x-tenant-id header. See Data Segregation.

With the full toolset, an agent can bootstrap a RAG stack end-to-end without touching the dashboard: create a namespace, request upload URLs, ingest documents, monitor the ingest job, then search and chat over the results.

Claude Code

claude mcp add --transport http agentset https://api.agentset.ai/mcp --header 'Authorization: Bearer agentset_xxx'

Claude Desktop and claude.ai

Go to Settings → Connectors → Add custom connector and enter https://api.agentset.ai/mcp as the server URL. The server authenticates with your API key as a bearer token.

Cursor

Add the server to .cursor/mcp.json in your project (or ~/.cursor/mcp.json globally):

{
  "mcpServers": {
    "agentset": {
      "url": "https://api.agentset.ai/mcp",
      "headers": {
        "Authorization": "Bearer agentset_xxx"
      }
    }
  }
}

Tools

Tool names are snake_case and map one-to-one to the REST API operations. Coverage by area:

Area	Tools
Namespaces	list_namespaces, get_namespace, create_namespace, update_namespace, delete_namespace
Uploads	create_upload_urls
Ingest jobs	list_ingest_jobs, get_ingest_job, create_ingest_job, reingest_job, delete_ingest_job
Documents	list_documents, get_document, delete_document, get_document_chunks_download_url, get_document_file_download_url
Retrieval	search, chat
Hosting	get_hosting, enable_hosting, update_hosting, disable_hosting
Custom domains	set_custom_domain, get_domain_status, remove_custom_domain
Webhooks	list_webhooks, get_webhook, create_webhook, update_webhook, delete_webhook, regenerate_webhook_secret
Organization	get_organization, update_organization, list_members
API keys	list_api_keys, create_api_key, revoke_api_key

Local MCP server

The @agentset/mcp package runs on your machine over stdio and exposes a single search tool scoped to one namespace. Use it when an agent only needs read access to your data.

Run it with your preferred package manager:

npm

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

Adding to Claude

To add the local MCP server to Claude, include the following configuration in your Claude settings:

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

Tips

Pass the namespace id as an environment variable:

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

Pass a custom tool description:

AGENTSET_API_KEY=your-api-key npx @agentset/mcp --ns your-namespace-id -d "Your custom tool description"

Pass a tenant id:

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

Next steps

• Agent-First Platform — How the REST API and MCP server cover the entire platform
• Search — Learn about search configuration options
• Data Segregation — Use tenant IDs for multi-tenant applications