GlockApps — Inbox Placement Testing MCP server

GlockApps MCP

Run real inbox placement tests from your AI assistant. Ask it to start a test, and it hands you the seed addresses to send to; ask again once you have sent, and it tells you who put you in the inbox, who put you in spam, and what to fix first.

The seed network is live mailboxes at Gmail, Outlook, Yahoo, AOL, Zoho, corporate Exchange and 30+ other providers.

https://api.glockapps.com/mcp

This repository is documentation and the registry manifest. The server itself is hosted by GlockApps — there is nothing here to install or run.

What you need

Account → API / MCP.

• A GlockApps account with API access on the plan. Your API key is on

• Test credits. Every test costs one, whether it is started here or in the GlockApps web app.

Quick connect

Claude.ai, Claude Desktop

Settings → Connectors → Add custom connector. Paste the address above and leave Client ID and Client Secret empty. Select Connect: a GlockApps page opens and asks for your API key once.

Claude Code

claude mcp add --transport http glockapps https://api.glockapps.com/mcp

Then /mcp in the session, and authenticate in the browser window it opens.

ChatGPT

Turn on developer mode (Settings → Security and login), then Plugins → + → Create app → Create MCP App. Paste the address, leave Authentication on OAuth, select Create, and paste your API key on the GlockApps page that opens. Developer mode is required while the connector is not yet in the ChatGPT directory.

Cursor, Windsurf, Continue and other clients

Clients that cannot do OAuth can send the API key directly. Copy a config from configs/ and put your key in it:

{
    "mcpServers": {
        "glockapps": {
            "url": "https://api.glockapps.com/mcp",
            "headers": { "Authorization": "Bearer YOUR_GLOCKAPPS_API_KEY" }
        }
    }
}

Full per-client instructions: docs/connect.md.

How a test works

The one thing to understand before you start: this connector cannot send your email. It gives you addresses to send to, and it reads what arrives.

seed address, with the marker included — either the X-API-Campaign-id header or the id: ... line in the body.

1. list_projects — resolves the project everything else needs.
2. createtest — costs 1 credit.** Returns the seed addresses and a marker.
3. You send. Send the campaign you want to test, from the platform you are testing, to every

1. waitfortest — results are usually usable about 10 minutes after sending.
2. get_results, then diagnose for what to fix, worst first.

A send without the marker never matches the test. Nothing arrives, the test never finishes, and the credit is still spent. If that happens, ask your assistant to delete the test: the credit comes back while no result has arrived.

Some providers accept mail slowly, so a test that has not "finished" after an hour is normal, not broken.

Things to ask it

• "Start an inbox placement test on my main project."
• "Has the test finished? What is the inbox rate?"
• "Why did this campaign land in spam?"
• "Compare this test with yesterday's. Did the fix work?"
• "Which providers are filtering us, and what should I change first?"

Tools

Fifteen tools, thirteen of them read-only. Full reference with parameters: docs/tools.md.

Security

Shortened. The full README is on GitHub.