Changelog
November 2026
Nov 18: Fewer false catch-alls. At mailbox providers whose servers accept every address, a check that cannot be completed now reads unknown, free and retried within the job, instead of catch-all. Addresses such a provider will not judge read unknown with reason no_signal.
Nov 18: No unknowns from waiting your turn. Some mailbox providers take only a few checks at a time. A check that has to wait for its turn now waits and then runs, instead of coming back unknown. POST /api/v1/verify/batch answers within about 80 seconds: an address whose check is still running then is listed under a new checking field, is not billed in that call, and returns its result when you call again with it. Match batch results to your list by email. Both SDKs carry the new field.
Nov 16: SPF checks in the API. GET /api/v1/domain/{domain}/audit and the MCP tool audit_domain_authentication read a domain’s SPF record the way receiving mail servers do: its syntax, the 10 DNS lookup and 2 void lookup limits counted through every include and redirect, and the policy for servers the record does not list. Each check costs 1 verification, and a check that could not finish is free. Test keys get a simulated answer for example.com. The free SPF Record Checker runs the same check in your browser, with no sign-up.
Nov 15: Domains that opt out. When a domain asks us not to check its addresses, its results come back unknown with reason domain_opt_out, and you are not charged.
Nov 14: Sandbox. Every account can now build and test against the API and the MCP server for free. Create a test key on the API keys page, or sign in from an AI app on the Free plan, and the test addresses at example.com return simulated results for every result type. A plan or a credit pack turns on live checks.
Nov 11: Connect AI apps by signing in. Add https://mcp.kavik.email/mcp to Claude, ChatGPT, Cursor, VS Code or another AI app, and approve it on a Kavik.email sign-in page. No API key to copy. The API keys page in your dashboard has one-click setup for Cursor and VS Code and lists the apps you connected.
Nov 5: Fewer unknowns. Addresses at more consumer mailbox providers, and on mail servers that answer slowly, now return valid or invalid instead of unknown. Domains whose mail servers are confirmed dead read invalid.
Nov 5: Sign in at kavik.email. The dashboard, sign-in and sign-up now live on kavik.email, with a new look and the same account, keys and usage.
Nov 3: USDT checkout. Plans and credit packs on kavik.email are paid in USDT on the TRON network (TRC-20), from the Billing tab in your dashboard.
October 2026
Oct 30: Slow checks over MCP. A check on a slow mail server no longer times out in your AI assistant. The tool replies checking while the check finishes in the background, and calling it again returns the verdict.
Oct 27: Fewer false invalids and false valids. Live mailboxes on strict mail servers are less likely to read invalid or risky by mistake, and consumer mailboxes need stronger evidence before an address reads valid. A reply we cannot read now returns unknown.
Oct 25: Catch-all results are free. A check that comes back catch-all is no longer billed, the same as unknown. It still appears in your history and exports. A confirmed answer on a catch-all domain, valid or invalid, is billed as before.
Oct 23: Free plan and credit packs. The 300 free checks are now a one-time allowance, one per person or company, at up to 50 a day in the dashboard. Free list uploads show a summary and up to 10 sample rows. Any plan or credit pack unlocks full results, the REST API and the MCP server, and the smallest pack is now 10,000 credits for $12.
Oct 14: Fewer catch-all results. A group of consumer domains that always read catchall now return valid when the mailbox is confirmed and invalid when it does not exist.
Oct 14: Cleaner results. A few undocumented diagnostic fields are no longer returned with a verification result or typed in the SDKs; the verdict, score, reason and mailbox flags already carried everything they said. Locked mailboxes now read is_locked.
Oct 5: Catch-all is its own verdict. Addresses on domains that accept everything now return catchall instead of risky, with the score (40 to 69) ordering how plausible the mailbox is. A catch-all address never reads invalid unless the mailbox was actually refused, and a provider-confirmed mailbox can still reach valid. Bulk exports, the dashboard, webhooks and both SDKs carry the new value.
Oct 4: Reason codes on every verdict. Every result now carries a reason field (for example catch_all, no_mx, greylisted, mailbox_not_found) explaining why a non-valid verdict landed where it did, across single, batch and bulk results, CSV exports and the dashboard.
September 2026
Sep 20: MCP server with OAuth. Verify addresses from AI assistants over the Model Context Protocol, with API keys or OAuth.
August 2026
Aug 30: Flat-rate pricing. $1 per 1,000 verifications on plans, $1.20 per 1,000 credit packs that never expire, full 28-check suite on every tier.