For years, if you asked an AI assistant "is my domain blacklisted?" it answered from memory or a web search - never a live DNS or blacklist query. That is starting to change.
What changed
The Model Context Protocol (MCP) is an open standard that lets an AI assistant call external tools during a conversation. Assistants including Claude, ChatGPT and Grok now support it, and Anthropic has launched a Connectors Directory where these tools are listed and connected in a click (OpenAI has a parallel app-submission process for ChatGPT). Instead of guessing, the assistant runs the real check on the tool's own infrastructure and answers from live data.
Why it matters for senders
Deliverability questions - blacklist status, SPF/DKIM/DMARC posture, spoofability, where a shortened link really goes, whether a message will land in spam - all depend on current state that is not in any training set. DNS zones are not crawled, and blacklists answer queries rather than publishing pages. A live tool call is what makes an assistant's answer trustworthy for the first time, with a link back to the full report.
Our part
Postbox Services is among the first email-deliverability tools in Anthropic's Connectors Directory: eight free checks (blacklists, spoofing / BEC grade, redirect tracing, a full spam test, raw .eml analysis, domain age and WHOIS), no account, connectable from Claude, ChatGPT, Grok and Cursor. See how it works and connect, or walk through eight real examples.