Find the Agent That's Holding Things Up

CanyonOS is a workflow-aware platform to build, scale, observe and manage multi-agent applications. It carries every call between your agents, so it knows which agent to scale, what each one costs and where each request spent its time.

Run it locallyRead the docs
YOUR LAPTOP OR YOUR OWN AWS ACCOUNTorchestratorworkflow.pyagent containerintent_agent.pyagent containermetrics_agent.pyagent containerrisk_agent.pyagent containeradvisor_agent.pyCanyonOSthe harness between your agentsyour model provider
TBD · scaling view screenshot

Scale.

Each agent works on a bounded number of requests, so work queues at the agent that's actually the bottleneck. CanyonOS adds or removes replicas of that agent, one at a time, within the limits in your scaling policy.

Docs guide

Assemble It Yourself, or Install One Platform

Assemble it yourselfCanyonOS
ScalingRules per agent, on a cluster you runScales the agent holding the others up, within your limits

What 0.2 Doesn't Do Yet

CanyonOS 0.2 is our first public release, and we ship fixes every week. Here's what it doesn't do today:

  • It hasn't run at large production scale.
  • It scales agent containers, not model capacity, so it won't speed up a slow model or lift a rate limit.