Skip to content
TextMeNotPowered byVetoAPI

Veto API · by TextMeNot

Junk text detection
in one call.

Send a text message, get back whether it is a scam, a cold sales pitch, a political text, brand marketing or a real message, with the words that decided it. Veto is TextMeNot's screening tech, the same family that screens texts in the TextMeNot app, as an API for your own product.

  • Under a millisecond per message
  • Message text is scored in memory and never stored
  • Five categories, so you decide what to filter
  • English and US Spanish

Try it

Paste any text message. This box needs no key and stores nothing.

Pricing

Monthly plans, billed through Stripe. Cancel or change plan any time from Manage billing. Taxes are added at checkout where they apply.

Free

$0 a month

100 checks a month

Enough to test it. No card.

Get a free key ›

Developer

$19 a month

25,000 checks a month

A side project or a small app in production.

Upgrade a key ›

Growth

$99 a month

200,000 checks a month

A messaging product with real traffic.

Upgrade a key ›

Business

$399 a month

2,000,000 checks a month

Carrier, CPaaS or platform volume.

Upgrade a key ›

Need more than 2 million a month, a contract or an invoice? Email support@textmenot.app.

Get a free key

100 checks a month to test with, no card. Your key appears right here.

By getting a key you agree to the API terms. We use your email only about your key and billing.

Upgrade or manage

Paste your key to move it to a paid plan, or to change plan, update your card or cancel.

Your key is only used to find your account; checkout and card details are handled by Stripe.

Docs

Check a message

curl https://textmenot.app/api/v1/messages/analyze \
  -H "Authorization: Bearer dk_live_..." \
  -H "Content-Type: application/json" \
  -d '{"text": "USPS: your package is on hold. Pay the $1.99 fee at usps-redelivery.top"}'
{
  "id": "req_ZOlAkWnhgGnP.scam.junk",
  "verdict": "junk",            // junk | review | ok
  "risk": 0.9874,               // 1 - P(real message)
  "risk_band": "high",          // high | medium | low
  "category": "scam",           // scam | junk | political | promo | ham
  "scores": { "ham": 0.0126, "scam": 0.9514, "junk": 0.0197, "political": 0.0008, "promo": 0.0155 },
  "reasons": ["fee", "hold", "package is", "on hold"],
  "model": "veto-2026-11",
  "latency_us": 190
}

verdict is our suggestion: junk when risk is 0.95 or more, review from 0.80, otherwise ok. Use scores to set your own cut-offs per category. promo is marketing from a real brand, junk is a cold pitch from someone the person never dealt with.

Several at once

Send up to 50 as {"messages": [{"id": "a1", "text": "..."}, ...]} and get {"results": [...]} back in the same order, each with your message_id. Each message counts as one check.

Other endpoints

  • GET /api/v1/usage: your plan, limit and checks per day this month.
  • POST /api/v1/feedback {"request_id": "req_...", "label": "ham"}: tell us an answer was wrong. Only the label is kept.
  • GET /api/v1/models: the model, its categories and thresholds. No key needed.

Errors and limits

  • 401 missing or unknown key, 400 no text, 413 over 4,000 characters, 429 monthly limit reached (resets on the 1st, UTC).
  • Every answer carries X-RateLimit-Limit and X-RateLimit-Remaining.

Privacy

The text you send is scored in memory and dropped when the answer is sent. We keep only counts per key per day, and for feedback, the two labels. Nothing you send trains the model.