Top up. Use. Stop any time.
No plans. No monthly commitment. Add $10 in credits, pay per call, and keep what you don't use — credits never expire.
Pay as you go·$10 minimum · Credits never expire · No plans
Pay as you goNo plansCredits never expire
Top up from $10. Use what you use.
No monthly plan. No commitment. Add credits once, and they stay on your account until you burn them. Every call is priced by what actually ran — cache hits cost pennies, full fan-outs cost the real LLM bill plus our margin. Transparent and bounded.
- $10 minimum top-up
- Credits never expire
- No card required to sign up
- Stop any time — no cancellation flow
Per-call pricing
Every call is billed by its shape.
Cheap by default, premium only when you ask for it.
| Call type | Price | $10 buys |
|---|---|---|
Cache hit Someone queried this same prompt recently; we served the cached answer. | $0.02 | 500 cache hits per $10 |
Single provider, no web search One LLM (ChatGPT, Claude, Gemini, or Perplexity), no live retrieval. | $0.05 | 200 calls per $10 |
Single provider + web search One LLM plus live web_search tool. Fresh citations. | $0.15 | 66 calls per $10 |
Multi-provider fanout 2–4 LLMs in parallel. No web search. Fast cross-comparison. | $0.25 | 40 calls per $10 |
Full fanout + web searchPlayground default 4 providers, live web search, brand extraction included. The headline product. | $0.75 | 13 calls per $10 |
All prices in USD. Brand extraction is bundled — never a separate line item. If every provider fails, you pay nothing for that call.
Enterprise
Need dedicated infrastructure, BYOK, or a custom contract?
We run Enterprise deals for teams at 150K+ calls/month, agencies with multi-tenant requirements, and companies with MSA/DPA/security-review needs. You will talk to Nikhil (founder) on the first call — no BDR layer, no discovery dance.
- · Custom per-call pricing
- · Dedicated cache infrastructure
- · Bring your own LLM keys (V2)
- · MSA, DPA, and custom security review
- · 99.99% uptime SLA
- · Dedicated solutions engineer
Every new account gets $1.00 in credit on signup, no card required, never expires. That's enough for ~50 quick-mode calls (the headline product) — long enough to wire MentionsAPI into your project and decide whether to top up.
Pay-as-you-go. Top up your balance ($5 minimum), every API call deducts in real time based on what actually ran (mode, partial-success refund, cache hit), top up again when you're low. No monthly subscriptions, no contracts, no overage bills — when your balance hits zero, the next call returns 402 until you top up again.
Most builders start with $20 and go from there. That covers 1,000 quick-mode calls or 80 perplexity_live calls — enough headroom to wire up a brand-monitor MVP. The minimum is $5 if you just want to evaluate.
No. Tokens are bundled into the per-call price. The number on the pricing table is the number you pay — no surprise OpenAI, Anthropic, or Google invoice hitting your card on the 1st of the month.
Default TTL is 24 hours, which is the right tradeoff for most brand-monitoring workloads: answer engines change their phrasing slowly enough that hourly polling mostly just grows your bill. The /v1/check?mode=perplexity_live path bypasses cache for ground-truth UI results.
No. Prompts and responses are never used to train models, and we never share data with third parties beyond the LLM provider you are querying. We log enough to debit and bill — nothing more — and our retention policy is linked from the DPA.
We return a partial response with the providers that did succeed plus an errors array describing each failure. If every provider fails, the call is free — you only pay for successful fan-outs (proportional refund applied automatically).
Stop any time by not topping up again — your remaining credits stay on your account indefinitely. Unused credits are non-refundable but also non-expiring, so there's no 'use it or lose it' pressure.
Both are on the Q3 2026 roadmap. Deep (multi-run variance + delta) needs the per-platform scraper concurrency story to hold; change-track (scheduled brand-rank watches) needs more reliable per-platform UI scraping. Both currently return 501 with a typed roadmap error so your code can branch deterministically. In the meantime, schedule mode:quick yourself or use POST /v1/watch with mode:quick.
Coming in V2. You will be able to attach your own OpenAI, Anthropic, Google, and Perplexity keys to bypass our bundled token cost, paying only the per-call infrastructure and extraction fee. Email [email protected] for early access.
Ready when you are.
Sign up free. Add $10 in credits. Start making API calls in under 60 seconds.