Model Context Protocol
Your agent can check a number before it texts it
Veriastra runs a remote MCP server. Point Claude, Cursor, or any client that speaks streamable HTTP at one URL and your agent can validate phones, verify email, score an IP, and screen a name against sanctions lists, in the middle of a conversation.
Connect it
There is no SDK to install and no command to run. The URL is the whole integration.
server url
https://veriastra.com/api/mcpAdd the URL
In Claude, open Settings, then Connectors, then Add custom connector. In Cursor, add it to your mcp.json as a server with a url and nothing else.
Sign in
Your client opens a browser window. You sign in to Veriastra and approve the connection. The client never sees an API key, and you can revoke it on its own from your dashboard afterwards.
Ask for something
The tools appear in your client. Calls run against your credit balance exactly as your own API keys do.
If your client cannot do the sign-in flow, set an Authorization header to Bearer <key> instead. A key beginning ol_test_ runs every engine and charges nothing, which is the honest way to try this before spending anything.
The 7 tools
Seven of our thirty-two endpoints, chosen rather than dumped. A tool list is context your agent spends on every turn, so a long one makes it worse at picking from it.
validate_phone1 crValidate a phone number
verify_email1 crVerify an email address
lookup_ip1 crLook up an IP address
lookup_domain1 crLook up a domain
screen_sanctions1 crScreen a name against sanctions lists
verify_business10 crVerify a business exists
fraud_score2 crScore a signup or order for fraud
Bulk jobs, the DNC registry endpoints and the scrapers are deliberately absent. An agent should not start a ten thousand row job on the strength of one sentence, and the registry endpoints carry obligations that belong to a signed-up caller rather than a chat turn. All thirty-two are on the API.
What the answers carry
The same thing the API returns, because the MCP server calls the same endpoints rather than a second copy of them. Which means the answer names its sources, states what it could not determine, and says why.
An email that cannot be proven comes back as unknown with the reason attached, greylisted or catch-all or port 25 blocked, not as a confident guess. An IP carries its BGP prefix, whether that prefix is RPKI valid, and the abuse contact from the registry that assigned it. A network we cannot classify is reported as unclassified rather than residential.
Connecting costs nothing. Calls spend credits at the rates above, from the same balance as the rest of your account. See pricing.