Create a sandbox key
Free, instant, shown once. Prefix ai92_test_ - it can never spend real credits.
AI92 for developers
One key.
150+ marketing operations.
Honest metering.
Contact discovery, creative generation, SEO, AI visibility, reviews and campaigns - one credit-priced REST API, a sandbox that never charges, and refunds on every failure.
Typical time to first call: under 3 minutes · Sandbox never charges · No card required
# Free read — your account & credit balance
curl https://api.ai92.ai/v1/account \
-H "Authorization: Bearer {{YOUR_SANDBOX_KEY}}"
HTTP/2 200
x-credits-consumed: 0
x-credits-remaining: 2516
x-ai92-sandbox: true
{ "account_id": "acct_9f2e17c4d8", "tier": "STARTER",
"credit_balance": 2516, "sandbox": true }From zero to 200 OK in four steps
Make a free read
Fetch your account or search contacts. Reads and overviews cost nothing, everywhere.
Run a paid action
Watch x-credits-consumed do the accounting in the response headers - sandbox meters but never charges.
Flip the prefix, go live
Same schema in production. Swap ai92_test_ for ai92_live_ and ship.
Fair metering - our contract with you
You only pay for what actually happened
- Reserve → refund. Async jobs reserve credits up front and refund them, exactly once, on failure.
- No charge on empty. Zero results or an inconclusive check? Not billed.
- No charge on our errors. A 5xx from us never costs you a credit.
- Pre-flight balance check. Insufficient credits returns a clean 402 before anything runs - no overdrafts.
- Per-key spending caps. Daily and monthly limits per key, plus IP allowlists and expiry.
All of this is documented behavior, verifiable in your ledger - not a pinky promise.
POST /v1/creative/generatesuccess−50 crPOST /v1/seo/auditjob failed upstreamrefunded +20 crGET /v1/contacts/searchread0 crPOST /v1/outreach/generatebalance too low402 · not run, not chargedThe whole marketing stack, one spec
150+ operations across the same services that power the AI92 dashboard. Prices live in the spec summaries - the reference, the rate card and this page are generated from it.
Contacts
7 ops · 2 freeSearch free · verify · reveal · enrich · bulk
Outreach
8 ops · 5 freeAI drafts · 1:1 sends · reply suggestions · newsletters
Creative
4 ops · 1 freeImages · video (beta) · brand-grounded copy
Campaigns
12 ops · 5 freeAI strategy · optimize · A/B test · budget allocation
SEO
7 ops · 2 freeAudits · keywords · backlinks · content optimization
AI Visibility
7 ops · 0 freeShare of voice in ChatGPT, Gemini, AI Overviews & more
Reviews & GBP
13 ops · 9 freeList · reply · request · Google Business posts
Markets & Intent
13 ops · 1 freeMarket reports · competitor intel · intent signals · visitors
Cloud data
5 ops · 2 freeBrand asset storage · audited data export
Webhooks
9 ops · 9 freeHMAC-signed · retries · dead-letter queue · replay
OAuth 2.1 + Brands
23 ops · 22 freeScoped access, rotating refresh tokens, per-brand isolation
Account & Billing
35 ops · 35 freeBalance · ledger · invoices · usage export - all by API
Documentation that respects you
26 pages of guides & reference, in the open
Quickstart, authentication, credits, rate limits, errors, idempotency, pagination, async jobs, webhooks, versioning, going-live - plus a full per-service reference and an interactive API explorer. No login required.
Open the docs API reference Download the OpenAPI spec Take the Build on AI92 course
Built for production
The boring guarantees, done properly
- Scoped API keys - hashed at rest, shown once, revocable instantly
- Per-key spending caps (daily/monthly) + IP allowlists + expiry
- Webhooks: HMAC signatures, timed retries, 30-day dead-letter queue, replay
- Idempotency keys with replay detection
- Versioning: additive by default, 12-month deprecation window
- 99.5% uptime SLA with published service credits
Credits from $29 · Reads free · Sandbox always free
Packs up to 55% off, subscriptions with a monthly credit bonus, and a per-action rate card generated from the spec.
Developer FAQ
How do I authenticate?
Send your API key as a Bearer token:
Authorization: Bearer YOUR_KEY. That is the only supported scheme - no X-API-Key header, no query-string keys. OAuth 2.1 with PKCE, scoped tokens and rotating refresh tokens is available for apps acting on behalf of users.Is there an SDK?
The API is plain REST with a fully-documented OpenAPI spec you can generate typed clients from, plus copy-paste examples in curl, Python and Node throughout the docs.
How does sandbox differ from production?
Same host, same schema, same endpoints, zero credit cost. Sandbox responses are metered so you can see exactly what production would charge, and carry the
x-ai92-sandbox: trueheader. When you are ready, switch the key prefix.What happens when I hit my rate limit?
You get a 429 with
Retry-Afterand standard rate-limit headers. Limits scale with your tier - from 10 to 50 requests per second.How do webhooks handle failures?
Deliveries are HMAC-signed and retried on a backoff schedule; exhausted deliveries land in a 30-day dead-letter queue you can inspect and replay from the dashboard or the API.
Can my agent or LLM read your docs?
Yes - the site ships an
llms.txtindex at ai92.ai/llms.txt, the docs are public with no login, and the full OpenAPI spec is downloadable, so AI assistants can consume everything directly.