Deliverability debugging is a perfect fit for agent workflows: tool-heavy, data-driven, bounded task graphs. We shipped an official MCP server so your agent can treat Inbox Check as a first-class tool.
What the MCP server exposes
run_placement_test— send a draft through the seed network, return per-provider placement.get_test_result— poll a previous test by token.monitor_domain— subscribe a domain to daily auth+blacklist+reputation checks.presend_analyze— spam-word/link intel/human-perception check on a draft before sending.list_providers— enumerate supported mailbox providers and seeds available.
Install
npm install -g ldm-inbox-check-mcp
# Or use directly via npx
npx ldm-inbox-check-mcp@latestClaude Desktop config
{
"mcpServers": {
"inbox-check": {
"command": "npx",
"args": ["-y", "ldm-inbox-check-mcp@latest"],
"env": { "INBOX_CHECK_TOKEN": "optional-for-higher-limits" }
}
}
}What agents do with it
- Deliverability triage. “My Gmail placement dropped — run a test and explain the cause” is a 30-second round trip.
- Pre-send review. An agent pipes your draft through
presend_analyzebefore the sequencer sends it. - Weekly digest. Scheduled agent runs all three tools, returns a 200-word health summary.
- Campaign QA. Agent fires a placement test for every new template variant, aborts promotion if below threshold.
MCP is free
No token required for basic usage — same rolling limits as the web UI. Token unlocks higher concurrency and longer retention. Start using it.
FAQ
Do I need a paid plan for the MCP server?
No. Anonymous use is rate-limited but fine for typical agent workflows. Token-authenticated use raises concurrency.
Is there an A2A agent endpoint too?
Yes — same tool set exposed over A2A. See our A2A protocol article.
What SDKs/runtimes work?
Anything that speaks MCP: Claude Desktop, Claude Code, Cursor, Continue, Zed, the Anthropic Agent SDK, OpenAI Responses API tool-use proxies, and custom stdio/http clients.
Related reading