Connect

Ask your DMARC data
anything.

Connect Claude, Cursor, or any MCP client to your workspace and ask the way you would ask a colleague. It reads your real reports and answers with clean tables, the same numbers you see in the app.

Your AI assistantExample

Is acme.com being spoofed? Was I protected last week?

Yes, and receivers report they blocked all of it. Here's last week for acme.com:

Spoofed messages blocked1,240
Delivered anyway0
Top sourceRussia

You're fully protected, receivers rejected every spoofed message before delivery.

What it can answer, and what it can do

Forty-six tools over your workspace: thirty-five that read, eleven that act, and five guided prompts to start from.

Read anything the dashboard shows

Which domains are enforcing, who sent as acme.com last week and from where, what p=reject would have stopped, the open actions, a diagnosis case, the lookalikes, Gmail's grades. Same loaders, same answers.

Act, if you let it

A write-enabled token can stage a one-click DNS fix, cancel one, name a sender, dismiss an action, ignore a lookalike, or add a webhook. Each write goes through the same safety net as a click in the app.

What it will not do

It never sees another workspace: a domain not in yours comes back as not found. A read token cannot write. A write re-checks your live role on every call, and nothing changes DNS without the five-minute hold and the 24-hour undo.

On the record

Every action an assistant takes is written to your audit log and the domain's history with the token that did it, by name, so a fix staged from a chat window reads exactly like one staged from the app. Tokens are per user, up to twenty per workspace, revocable in one click.

Beyond the tools, the server offers guided starters (investigate spoofing, plan enforcement, a weekly review, a pre-add check, explain these headers), browsable workspace and per-domain views, and autocomplete for your own domain names, so an assistant can find its way without you spelling everything out.

Two steps, any client

1

Create a token

Settings, Integrations, MCP. Name it for the machine it lives on. Read-only by default; tick write access only if you want the assistant to act.

2

Point your client at the server

A remote server at mcp.trustyourinbox.com with a bearer token. Cursor and the Claude desktop app take a short JSON block; any client that speaks Streamable HTTP works.

3

Start with list_my_domains

The server tells the assistant where to begin. From there, ask in plain English and follow the guided prompts when you want a structured review.

The MCP server is included on every plan. One tool, creating a webhook endpoint, follows the webhook plan line.

Ask the first question
tonight.

Add a domain, mint a token, paste it into your client. The first answer is a table of what sent as you this week.