MCP Server Integration

The AIdenID MCP (Model Context Protocol) server lets LLMs use supporting identity, extraction, and lifecycle tools. MCP is useful for issuer and demo workflows; it does not sit on the per-request clearance hot path.

What is MCP?

The Model Context Protocol is a standard for giving LLMs access to external tools. When you configure an MCP server, the LLM can call AIdenID functions as part of its reasoning loop — no custom integration code required.

Setup

Add the AIdenID MCP server to your LLM configuration. The exact configuration depends on your client, but the pattern is the same:

Claude Desktop / Claude Code

{
  "mcpServers": {
    "aidenid": {
      "command": "npx",
      "args": ["-y", "@aidenid/mcp-server"],
      "env": {
        "AIDENID_API_KEY": "aid_your_api_key",
        "AIDENID_ORG_ID": "org_abc123",
        "AIDENID_PROJECT_ID": "proj_def456"
      }
    }
  }
}

Environment variables

Variable	Required	Description
`AIDENID_API_KEY`	Yes	Your AIdenID API key
`AIDENID_ORG_ID`	Yes	Your organization identifier
`AIDENID_PROJECT_ID`	Yes	The project to scope operations to
`AIDENID_API_BASE`	No	API base URL (defaults to `https://api.aidenid.com`)

Available tools

Once configured, the LLM has access to the following tools:

create_identity

Provision a new disposable email identity.

Parameter	Type	Required	Description
`label`	string	Yes	Human-readable label
`ttl_hours`	integer	No	Time-to-live in hours (default: 24)
`domain`	string	No	Custom domain for the inbox

get_identity

Retrieve identity details including status, email, and metadata.

Parameter	Type	Required	Description
`identity_id`	string	Yes	The identity ID to retrieve

list_identities

List identities in the current project with optional filtering.

Parameter	Type	Required	Description
`status`	string	No	Filter by status
`limit`	integer	No	Max results (default: 20)

get_latest_extraction

Get the most recent extraction result for an identity.

Parameter	Type	Required	Description
`identity_id`	string	Yes	The identity to check for extractions

extend_identity

Extend the TTL of an active identity.

Parameter	Type	Required	Description
`identity_id`	string	Yes	The identity to extend
`additional_hours`	integer	Yes	Hours to add

squash_identity

Revoke an identity and deactivate its inbox.

Parameter	Type	Required	Description
`identity_id`	string	Yes	The identity to squash

create_webhook

Parameter	Type	Required	Description
`url`	string	Yes	HTTPS endpoint URL
`events`	string[]	Yes	Event types to subscribe to

test_auth_flow

Run an auth flow reliability test against a target URL.

Parameter	Type	Required	Description
`flow_type`	string	Yes	`password_reset` or `invite`
`target_url`	string	Yes	URL to test
`identity_id`	string	Yes	Identity to use for the test

Example conversation

Here is how an LLM might use AIdenID tools in a conversation:

User: Run the authorized signup test on example.com using a disposable email.

LLM: I'll create a disposable identity for the authorized test flow.

[Tool call: create_identity]
  label: "example-signup"
  ttl_hours: 1

[Tool result]
  id: "ident_a1b2c3d4e5"
  email: "a1b2c3d4e5@inbox.aidenid.com"

LLM: I've created the identity. Let me use the email in the test flow...
     Now I'll check whether extraction evidence is ready.

[Tool call: get_latest_extraction]
  identity_id: "ident_a1b2c3d4e5"

[Tool result]
  type: "otp"
  has_secret: true
  redacted_value: "***"
  confidence: 0.99

LLM: Extraction evidence is ready. The raw secret must stay in the secure
     server-side runtime path. Let me clean up the identity.

[Tool call: squash_identity]
  identity_id: "ident_a1b2c3d4e5"

Security considerations

API key access. The MCP server has full access to your API key scopes. Use a key with minimal permissions.
Environment isolation. Use a separate project for MCP-driven operations so they do not interfere with production identities.
TTL hygiene. Set short TTLs when using MCP tools to minimize active identity window.
No token passthrough. Do not pass raw OTPs, magic links, or upstream bearer credentials through LLM-visible context.