Connect to the Anzenna MCP Server

Anzenna hosts a Model Context Protocol (MCP) server that exposes your Anzenna security data (people, devices, DLP alerts, investigations, login events, and more) as tools that an AI assistant can call. Point any MCP-capable client (such as Claude) at the server with an Anzenna API key, and it can query and act on your data through natural language.

Access follows your API key

The MCP server authenticates with an Anzenna API key and performs every action as that key. What the assistant can see and do is governed by the key's permissions, so scope the key to least privilege before connecting. See

API Keys

to create and scope one.

Connection Details

Setting	Value
Endpoint	`https://mcp.anzenna.ai/mcp`
Transport	Streamable HTTP
Authentication	Bearer token: your Anzenna API key ( `anzak1-…`)
Auth header	`Authorization: Bearer <your-api-token>`

Prerequisites

An Anzenna API key, scoped to what the assistant needs (see
API Keys
)
An MCP-capable client that supports remote (HTTP) servers

Connect a Client

Claude Code (CLI)

Add Anzenna as a remote MCP server, passing your API key in the Authorization header:

claude mcp add --transport http anzenna https://mcp.anzenna.ai/mcp \
  --header "Authorization: Bearer $ANZENNA_API_TOKEN"

Generic MCP client (JSON config)

Most MCP clients accept a JSON entry describing the remote server. Add an entry like the following, substituting your API key:

{
  "mcpServers": {
    "anzenna": {
      "type": "http",
      "url": "https://mcp.anzenna.ai/mcp",
      "headers": {
        "Authorization": "Bearer <your-api-token>"
      }
    }
  }
}

note

If your client only supports local (stdio) MCP servers, use a stdio-to-HTTP bridge to reach the remote endpoint, passing the same bearer token. Consult your client's documentation for how it configures remote servers.

What You Can Do

The server exposes Anzenna's API as MCP tools. Most are read-only queries over your security data; a few perform actions. Examples:

Category	Example tools
Query data (read-only)	`query_people`, `query_devices`, `query_dlp_alerts`, `query_investigations`, `query_login_events`, `query_oauth_applications`, `query_shadow_ai_instances`
Look up a record	`get_person_by_id`, `get_device_by_id`
Advanced / free-form	`run_advanced_query`
Actions (require write access)	`execute_remediation`, `create_allowlist`, `dismiss_investigation`

Read-only tools are annotated as such, so clients that surface tool safety hints will show them as non-destructive. Action tools take effect immediately and require an API key with the corresponding write permission.

Keep your token secret

Your API key grants access to Anzenna data through the assistant. Store it in a secrets manager, never commit it to source control, and revoke it from

API Keys

if it is exposed.

Need help? Contact

Anzenna Support

for assistance.

Connection Details​

Prerequisites​

Connect a Client​

Claude Code (CLI)​

Generic MCP client (JSON config)​

What You Can Do​