Pricing
We are in early access and priced like it. No seat minimums, no token markup, no surprises — and the early-access rate is the rate you keep.
For one team getting its AI usage under control.
For companies running AI across departments.
For regulated and large organizations.
Routed model spend is billed at provider cost — Pong never marks up tokens. On paid tiers, a 2–4% governance fee applies to routed spend. That fee is the entire routing-side business model: our margin never grows by making your token bill grow.
Because a router that profits from your token volume can’t be trusted to minimize it. Pong’s incentives have to point in the same direction as yours, so we charge for governance, not for volume: model spend passes through at provider cost, and a small governance fee on routed spend funds the policy engine, memory, and audit infrastructure. Neutrality is the product.
ZDR means the model provider does not store your prompts or outputs after the request completes — which also puts you outside the blast radius when courts compel providers to preserve conversations. Pong enforces it at route time: a ZDR-required request is only ever sent to a verified ZDR endpoint, its fallback chain contains only other ZDR endpoints, and the retention terms in force are recorded on the request’s audit event. It is enforced per request, not promised per contract.
Yes — your data, exportable, always. Organization, department, and user memory can be exported in full, with provenance, on every tier. Memory portability across models is the reason Pong exists; locking you in would defeat the point.
BYOK routes model traffic through your own provider keys: your negotiated rates, your contracts, your ZDR agreements — and Pong never marks up that spend. Hosted keys are the fastest way to start in early access. Routing, policy, memory, and audit behave identically either way, and you can switch to BYOK at any time.