Open source · AGPL-3.0

The One Platform for Scaling your Agentic Apps

Write agents in plain Python · Deploy with one command · Auto-scale just the right blocks · Observe with no additional code

curl -fsSL https://raw.githubusercontent.com/CanyonCodeCoreAI/canyonos/main/packages/cli/install.sh | sh
canyonos quickstart
Your WorkflowsAgentic Apps
CanyonOSThe Layer for Scaling your Workflows
Your computeYour Laptop, AWS or Kubernetes, Any OpenAI-Compatible Provider

See How It Works

Scaling your workflows - Made Easy.

Plain Python

No SDK, no decorators, no graph to define.

How setup works One-command deploy

Deploy to your AWS account or a Kubernetes cluster from the same CLI.

How deployment works Independent scaling

Scale one agent, not the whole app.

How scaling works Observability without tracing code

Get unprecedented per-workflow, per-agent, per-query insights.

How observability works Policy-based governance

Spend and model rules per agent in a policy file, changed without a redeploy.

How governance works

Your Cloud, Your Models, Your Data

CanyonOS is cloud-ready and model-agnostic. Workflows, prompts and data stay in your environment. The CLI sends anonymous usage data, and one setting turns it off.

CloudYour laptop, your own AWS account or your own Kubernetes cluster.
ModelsAny provider with an OpenAI-compatible API, your own keys.
TelemetryOpenTelemetry-compatible traces, metrics and logs, exportable to Datadog, Grafana, Prometheus, Langfuse or Arize Phoenix.

FAQ

Do I need an account to run this?

No. Install it and run the Hello World quickstart on your laptop; it needs no API key. Bring your own model provider when you deploy your own workflow.

Do I need Kubernetes?

No. CanyonOS runs on Docker. If you already run Kubernetes, it can deploy there too, with an operator and a Helm chart.

Do I have to rewrite my agents?

Yes, into plain Python, one file per agent. The optional helper converts a copy of a LangChain or LangGraph project for you to check.

Which model providers does it work with?

Any provider with an OpenAI-compatible API. Most providers offer one, and so do common servers for models you host yourself.

Can I keep my observability tools?

Yes. CanyonOS exports traces and metrics to them and adds cost and queue time per agent.

Is it production-ready?

It's 0.2, our first public release, and an active project with weekly releases. It hasn't run at large production scale yet. The current limits are listed in the README.

What does AGPL-3.0 mean for my company?

You can download, run and change CanyonOS for free. If you modify it and offer it to others over a network, AGPL requires you to make the source of those changes available to them.

What does the CLI send home?

An anonymous install ID, the command, whether it worked and the CLI version. No code, prompts or model data, and one setting turns it off.

Why not assemble this from tools I already use?

You can, as a platform project. CanyonOS gives you per-agent scaling, tracing and cost in one install, because it carries every call.

How does Canyon Code make money?

CanyonOS is free and open source. We plan to offer paid services around it later.