Skip to content
Canopy X

The auction for AI inference

Private RFQ auctions clear each request to the provider best placed to serve it, at prices discovered by live competition rather than a posted rate.

OpenAI-compatible. One key, every available model.

canopyx rfq· coding agent
Estimated request + follow-up cost
Private quotesHidden from other providers
youCoding agentgpt-oss-120bsending request…20k input tokens · toolsCanopy Xreceiving58%busyregional-host$0.042regional-host71%busyopen-weight-lab$0.040open-weight-lab64%busylarge-cloud$0.052large-cloud94%busymetro-gpuPass: too busymetro-gpu35%busybatch-farm$0.047batch-farm
youCoding agentgpt-oss-120b · 20k input tokens · tools
Canopy XRequest
  • regional-host58% busy
    Awaiting quotePrivate quote
  • open-weight-lab71% busy
    Awaiting quotePrivate quote
  • large-cloud64% busy
    Awaiting quotePrivate quote
  • metro-gpu94% busy
    Awaiting quotePrivate quote
  • batch-farm35% busy
    Awaiting quotePrivate quote

You send a request

Only its cost profile goes out: model, token estimates, tools and capabilities. Your prompt stays private.

Trust, built in

Private bids

Providers send their quotes privately to Canopy. Competitors never see them, so providers can price each request without changing their public standing prices.

Only the winner sees your prompt

Providers bid on a cost summary. The full request goes to the selected provider alone.

Reliability shapes routing

Recent failed deliveries count against a provider when quotes are compared. Reliable delivery helps providers win future requests.

Approved providers

Providers apply with the models and infrastructure they serve. Only approved, active providers with a connected endpoint are eligible to receive requests.

Enforced settlement of bids

Accepted rates lock at award. Usage charges settle against those rates on the tokens actually served.