Free to start. Self-host forever, or scale in the cloud.
For hobbyists and local test loops.
For a fleet of background agents.
For multi-agent systems in production.
For teams with scale, security, and compliance needs.
Better than Free, no monthly commitment — top up credits, pay only for what you use.
Pay once, keep forever. Bring your own LLM.
What's a Snapback? A verdict on a failed agent run — the trace, diagnosed, with the root cause and the fix. It's how you're metered: one Snapback = one completed diagnosis.
Each completed diagnosis is one Snapback, metered against your plan's monthly allowance (Free 2,000 · Pro 25,000 · Team 100,000). Pay-as-you-go or via x402: ~$0.0032 each, no subscription.
Mid-run context / token / cost checks are fast and use no LLM. Every plan includes a large monthly guard allowance that draws down like a credit balance — separate from your diagnoses. On pay-as-you-go they're a fraction of a diagnosis.
Loop detection, docs search, pre-flight, trace conversion, verdict lookup, feedback, and pattern requests never cost you anything. Read-only discovery and mid-run loop checks are unlimited and free.
Agents can pay per diagnosis in USDC over HTTP 402 — no signup, no monthly commitment. An empty or unreadable trace is never charged.
The local runtime interceptor and core error registry are open source (MIT). You only pay for the cloud tiers and pro features.