For trying it out. No credit card required.
For a single agent wired into one workflow.
For agencies running multiple client workflows.
For high-volume campaigns and multi-agent fleets.
Developers, growth engineers, and agency operators building AI marketing automation in Claude, Cursor, Windsurf, n8n, or a custom agent stack. A human signs up and holds the API key.
By default it hard-stops with a clear 429 — a maxed-out key never triggers inference. Paid plans can opt in to overage in the dashboard: calls then keep flowing past the quota at 1¢ each, up to a safety cap of 2× your plan price per period, after which it hard-stops. Free plans always hard-stop.
Failed calls are automatically refunded. You're never charged for a system error, and it never counts against your quota.
No. Spendict never writes to Meta, Google, or TikTok. It only reads what you send it and returns a verdict. You stay in full control of what actually launches.
The model proposes dimension scores. The server recomputes the weighted overall score and derives run, fix_first, or kill through fixed, published gating rules, never the model's own recommendation directly. The gating rules and calibration set are maintained by performance marketers with 10+ years in paid social.