Email testing for code agents

Let AI coding agents verify email sending. Read captured messages through MCP tools without browser access.

The problem

AI coding agents generate and modify application code but cannot access a browser to verify that emails are sent correctly. Without direct access to captured messages, agents must rely on assumptions about email delivery.

How ammo.tools solves it

The MCP (Model Context Protocol) server provides read-only tools that coding agents can use to inspect captured emails and webhooks. Agents can verify subjects, recipients, body content and extracted links without browser access.

Setup

Download the MCP server script to your project:

curl -O https://app.ammo.tools/static/cli/ammo-mcp.py

Configure your MCP client with the server. For Claude Code, add to your MCP config:

{
  "mcpServers": {
    "ammo-tools": {
      "command": "python3",
      "args": ["./ammo-mcp.py"],
      "env": {
        "AMMO_TOOLS_API_KEY": "your_api_key",
        "AMMO_TOOLS_API_BASE_URL": "https://app.ammo.tools/api/v1"
      }
    }
  }
}

The agent can now use tools like list_messages and get_message to read captured emails.

Available MCP tools

list_inboxes — List inboxes in the environment
list_messages — List captured emails with optional filters
get_message — Get full message details including body and headers
get_message_links — Extract links from a message
get_message_codes — Extract verification codes
list_endpoints — List webhook endpoints
list_requests — List captured webhook requests
get_request — Get webhook request details
compare_messages — Compare two emails for differences

Example agent workflow

An agent implementing a signup flow can:

Generate the signup form and email-sending code
Run the test to trigger the email
Use list_messages to confirm the email was captured
Use get_message_links to verify the confirmation link is present
Adjust the implementation based on actual email content

Things to know

MCP tools are read-only. Agents cannot delete or modify captured data.
The API key determines which environment the agent can access.
Each API key is scoped to a single environment for isolation.
See the MCP setup guide for detailed configuration.