Endpoints

Method	Path	What it does	Billed
`POST`	`/v1/check`	Check one address	1 email
`GET`	`/v1/check?email=...`	Same, using the query string	1 email
`POST`	`/v1/check/bulk`	Check up to 1,000 addresses	1 per address answered
`POST`	`/v1/feedback`	Report what happened to a checked address	Free
`GET`	`/v1/model`	Live accuracy of the model	Free

`POST /v1/check`

Field	Type	Required	Description
`email`	string	Yes	The address to check. Max 254 characters.
`ip`	string	No	The user's public IP (IPv4 or IPv6). Adds cloud-server, Tor and same-IP checks. Never stored.
`explain`	boolean	No	`true` adds an `evidence` list showing how each signal moved the score.

Prefer POST over GET: addresses then stay out of URLs and server logs. The GET version takes the same fields as query parameters.

`POST /v1/check/bulk`

{ "emails": ["a.lee@gmail.com", "not-an-email", "test@mailinator.com"] }

Returns one entry per input, in the same order. An invalid input returns an error entry instead of failing the whole request:

{
  "results": [
    { "email": "a.lee@gmail.com", "human_score": 94, "verdict": "allow", "...": "..." },
    { "input": "not-an-email", "error": "not a valid email address" },
    { "email": "test@mailinator.com", "human_score": 0, "verdict": "block", "...": "..." }
  ]
}

Addresses are checked in parallel, and addresses at the same domain share one lookup, so a bulk call is much faster than many single calls.

`POST /v1/feedback`

{ "check_id": "chk_AXDmekmC35NL1mpGtCtC...", "outcome": "fake" }

Or skip the check_id and send the address. It labels your latest check of that inbox:

{ "email": "maria.santos@gmail.com", "outcome": "confirmed" }

Returns 202 Accepted. See Teach the model with feedback.

`GET /v1/model`

Shows how accurate the model is, measured on real feedback before the model learned from each example. Example response (your numbers will differ):

{
  "mode": "auto",
  "serving": "baseline",
  "healthy": false,
  "learned_version": 12,
  "feedback_examples": 140,
  "customers": 3,
  "baseline": { "accuracy": 0.91, "log_loss": 0.27, "recent_log_loss": 0.25 },
  "learned":  { "accuracy": 0.93, "log_loss": 0.22, "recent_log_loss": 0.21 },
  "recommendation": "keep collecting feedback"
}