Connect your AI

Tirada has an MCP server. Connect your assistant, whether it's Claude, ChatGPT or another one, and it can check why an email didn't arrive, whether your domain is set up right or what happened with an import, and do in your account what you allow it to. Sending is the exception: a person confirms every send in the dashboard.

Before you start

You need a Tirada account and a computer. When you connect, your assistant sends you to the Tirada dashboard to sign in and approve access, and the dashboard runs on a computer. The server address is the same for every client:

https://mcp.tirada.app/mcp

Claude and ChatGPT rename their menus now and then. If you can't find one, search their settings for "connector" or "MCP".

Claude

1. On claude.ai, open Settings and go to Connectors.
2. Click "Add custom connector", name it Tirada and paste the URL https://mcp.tirada.app/mcp.
3. Click Connect. The Tirada dashboard opens: sign in, pick the workspace and publications, review the permissions and approve.
4. Back in Claude, turn the connector on in the conversation and ask something like "Is my newsletter's domain set up correctly?".

In Claude Code it's one line. Then type /mcp inside Claude Code to sign in:

claude mcp add --transport http tirada https://mcp.tirada.app/mcp

ChatGPT

1. In ChatGPT, open Settings → Apps & Connectors → Advanced settings and turn on developer mode.
2. Create a connector with the URL https://mcp.tirada.app/mcp and OAuth authentication.
3. When you connect it, ChatGPT takes you to the Tirada dashboard to sign in and approve the permissions.

Other MCP clients

Any client that speaks MCP over Streamable HTTP with OAuth 2.1 and dynamic client registration works. The client registers itself, opens the Tirada dashboard for you to approve, and gets a token that expires every hour and renews on its own. If your client wants a JSON config, this is usually enough:

{ "mcpServers": { "tirada": { "url": "https://mcp.tirada.app/mcp" } } }

The OAuth metadata lives at https://mcp.tirada.app/.well-known/oauth-protected-resource.

What your agent can do

Today it has eight diagnostic tools, all read-only:

• check your domain: which DNS records we expect, which we found and what's missing;
• follow one reader's email in a send: whether it went out, bounced or is suppressed;
• count a send's bounces and complaints against your publication's average and the thresholds that would pause it;
• read an import report: how many got in and why the rest were left out;
• check a post before you send it;
• explain why a send is paused and what unblocks it;
• show the state of your Stripe account and payments;
• search Tirada's documentation.

It can also use the operations of the public API you approve on the permissions screen, such as reading your posts or preparing a draft, and never more than your role allows. If someone removes your access to the workspace, the agent loses it with you.

What it can't do

No send goes out until a person confirms it in the Tirada dashboard. The agent can leave a send ready; only you can confirm it. It also can't publish posts, create API keys or webhooks, export data, change your domain or payments, invite anyone or unsubscribe a reader.

Your readers' emails reach it masked (a***@domain.com) and it can read at most 5,000 subscriber records a day. Every connection shows up in the dashboard under Ajustes → Agentes conectados (Settings → Connected agents; the dashboard is in Spanish for now), and revoking it cuts access on the spot.

Which plans

Diagnostics with your own AI come with every plan, Lite at €5 a month included. Basic, at €15, adds an AI assistant inside Tirada that knows your account. See pricing.