Let Claude, ChatGPT, Cursor or another MCP client verify addresses for you: create an agent, add the PricklyMails server, approve it, and revoke it anytime.
An AI assistant can verify addresses for you in the middle of a conversation: you ask in plain words, and it calls PricklyMails.
It connects to the PricklyMails MCP server, at https://mcp.pricklymails.com/mcp, through an agent you create in the dashboard.
No API key is pasted anywhere: you approve the agent once, and revoking it cuts the assistant off.
In the sidebar, under Developer, open Agents and click New agent.
Name it after the assistant that will use it, such as Claude; the Description is optional.
Under Scopes, tick what the assistant may do, or start from one of the Quick presets: Verification agent, Verification and batches, Read-only or Full access.
Click Create agent.
Claude: open Customize then Connectors, click + Add, then Add custom connector, and enter the URL; the Claude page has every step, Claude Code included.
ChatGPT: on a Business, Enterprise or Edu workspace, on the web, open Plugins, select the plus button, then Add custom MCP server, enter the URL and choose OAuth; the ChatGPT page has every step.
Cursor: add the server to ~/.cursor/mcp.json, or use the install button on the Cursor page.
Another MCP client: add a remote MCP server with the URL, and the client signs in through OAuth, with no key to paste.
The assistant opens PricklyMails in your browser: sign in to your account.
On the Connect an agent page, check the address you will be sent back to.
Under Choose which agent to authorize, pick the agent you created, and read what it can do.
Click Authorize.
Warning
When the page says it does not recognize the application, or that the authorization goes to a program on your computer, continue only if you started the connection yourself.
Otherwise, click Deny.
Within the scopes of its agent, the assistant can verify one address or a whole list, read the results and their reasons, check your balance and usage, and manage your API keys and webhooks.
Each address it verifies runs every check your plan allows, disposable addresses included, and costs the same tokens as a verification of every check from the dashboard (Tokens and plans).
Each verification spends tokens from your account, like the same request through the API: see how tokens work.
How many agents you can create, and which scopes you can grant, depend on your plan: a scope marked Upgrade needs a higher one.
See the plans.
PricklyMails receives only what a tool call needs, such as the address to verify, never your conversation, and it does not store the address: see Your data.
Every tool the server offers is listed in the MCP server docs.
In Agents, click the agent's row to open Agent details, click Revoke, then Revoke agent.
The assistant loses access immediately, and this cannot be undone.
You can also remove PricklyMails from the assistant itself:
Claude: open the connector in Customize then Connectors and click Disconnect.
ChatGPT: in Settings then Plugins, open PricklyMails, then its three-dot menu, and choose Disconnect.
Cursor: turn the server off in Customize then MCPs, or delete its entry from mcp.json.
An agent's scopes cannot be changed once it exists.
To give an assistant more or less access, create a new agent and revoke the old one.
Verify a list in one batch: upload or paste it, follow its progress, read and download the report, and see what happens to duplicates and invalid lines.
How billing works: tokens spent check by check, what each check costs, plans and top-ups, tokens that never expire, and what is refunded or never billed.
Set a strong password, turn on two-factor authentication, sign out the devices you do not recognize, and know which security emails your account sends you.
What happens to the addresses you verify, what the shared cache and your own browser keep, and what remains once you delete your account, and for how long.
Check your list before your next send.
1,000 tokens free when you sign up, no credit card. After that, you only pay for answers.