Questions teams ask before adopting SpendPilot.

SpendPilot tracks, alerts, and optimises AI agent-fleet spend in real time across OpenAI, Anthropic, and other supported providers. It watches every API call, detects anomalies, reroutes traffic away from runaway agents, fires kill switches the moment a threshold is breached, and delivers a daily CFO-style report. Unlike cost dashboards that only show you the problem, SpendPilot takes action on it. Learn more on the How It Works page.

SpendPilot is built for engineering teams running 10+ AI agents, AI-ops leads who own the fleet's bill, platform and infrastructure teams that already gatekeeper model access, and the finance co-pilots who need spend broken down by team, model, and agent. Hobby projects and pure end-developer tinkering are out of scope — SpendPilot is most useful the moment spend becomes a multi-thousand-dollar line item a stakeholder wants explained. See how Scale fits a real rollout on the Pricing page.

Starter covers 1 API key and 1 provider for teams proving out the workflow. Pro goes up to 5 API keys across 3 providers with unlimited agent fleet monitoring on each — most growing fleets land here. Scale supports 10+ API keys across every supported provider, autonomous rerouting, spend forecasting, custom thresholds, and per-agent budgets; Scale includes autoscale behaviour that cleanly absorbs new agents without manual reconfiguration.

API keys are encrypted at rest; you manage them after signing in. SpendPilot is read-only at the provider-account level — it never proxies your requests, so there is no SDK or request-side surface to attack. Scale customers get SOC-style access controls, named-account support, audit-log export, and SSO-ready provisioning. We never read your prompts or responses; only spend signals.

Under five minutes. Zero code changes, no SDK install, and no proxy configuration. You connect your AI provider API keys and SpendPilot starts monitoring immediately. Your agents continue calling OpenAI, Anthropic, and other providers exactly as they do today — SpendPilot observes and acts at the account level, not the request level. Read the mechanics on the Alerts page.

A cost dashboard shows you what already happened. SpendPilot acts on what's about to happen. Dashboards give your team a chart and a Slack notification; SpendPilot reroutes traffic between providers, throttles misbehaving agents, and fires kill switches the moment a threshold is breached. If you only need a chart, a dashboard is fine. If you need the bill to stop growing, see the side-by-side on the Compare page.

Three tiers. Starter is free for 3 days with no credit card required (1 API key, 1 provider, email alerts). Pro is $49/mo and unlocks up to 5 API keys, 3 providers, real-time anomaly detection and kill switches, autonomous multi-provider routing, daily CFO reports, and unlimited agent fleet monitoring. Scale is $199/mo for 10+ keys across every supported provider, autonomous rerouting, spend forecasting, custom thresholds, per-agent budgets, dedicated onboarding, SSO-ready provisioning, and a 4-hour support SLA. Full breakdown on the Pricing page.

Yes — Stripe powers SpendPilot's own checkout and subscription webhooks, so you already interact with Stripe in the billing flow. SpendPilot is not, however, a spend-data source for Stripe charges; those flow through your own Stripe account as usual. Out-of-the-box data-source adapters ship for OpenAI, Anthropic, Gemini, and Cohere. Supabase rows are not a spend data source; if you are tracking usage there, the best path forward is to forward aggregates through /api/track/spend-events or a Stripe webhook into SpendPilot.