Cloud World Model documentation

Start with a simulation

Create a virtual infrastructure graph, advance the model and inspect simulated metrics. The walkthrough includes copyable requests and response shapes; it is the best starting point for a first API call.

How this differs from chaos tools, FinOps calculators, and monitoring

Chaos and fault-injection tools affect real systems; CWM uses simulated chaos and recommends later production-capable validation such as AWS FIS or Gremlin. FinOps calculators estimate prices from SKUs or IaC diffs; CWM also models latency, errors, and resilience under traffic and failure, but its outputs are estimates, not invoices. Monitoring observes live telemetry after deployment; CWM is a pre-deployment and pre-change world model, not an observability or operations console.

MCP quickstart

Connect an AI assistant through the MCP server. The quickstart covers connection configuration and a first tool call.

Agent workflows

For agent-oriented APIs, examples and tool descriptions, read the agent reference. Pair it with the simulation walkthrough before automating changes.

RL environments

Use a linked simulation as an episodic training environment. The environment reference documents create, step, observation, reset, actions and reward fields; read it before designing an autoscaling policy.

Chaos and scenarios

Browse modeled failure scenarios and run chaos jobs against a simulation, not live infrastructure. The use-case guide describes asynchronous jobs and result retrieval.

Accuracy evidence

Read benchmark methodology, reference comparisons and metric-specific limitations before using estimates to make a deployment decision.

Model fidelity

Check which behaviors the simulator models and where documentation-backed assumptions or unsupported behavior limit transfer to production.

REST API reference

• Interactive OpenAPI reference: /api-docs/
• Machine-readable simulator description: /api/description

Browse endpoint schemas, authentication and request/response definitions in the interactive API reference. For a machine-readable description of simulator capabilities, see the simulator description endpoint.

Provider coverage

Check which cloud services, instance-size catalogs and billing models are represented for AWS, GCP, Azure, OCI and DigitalOcean. Each service is marked as measured or extrapolated so you can distinguish direct benchmark validation from model-derived behavior before building a comparison.

Multi-cloud burst traffic report

The report presents a specific simulated workload across five cloud providers under baseline and burst traffic. Its output is scenario-specific and should not be treated as a universal provider ranking, an invoice or a production measurement.

Guide: pre-provision multi-cloud simulation

Evaluate greenfield designs and proposed changes to existing systems before deploying or changing infrastructure; define a virtual baseline, inspect modeled cost and latency, then review the evidence and assumptions before comparing deployment options.

Guide: RL autoscaling environments

Follow a valid environment-create and step example, learn how observations, nested actions and reward components fit together, then review evaluation and sim-to-real limits before using a policy beyond the simulator.

RL environments overview

Guide: chaos without production changes

Run a modeled failure experiment, inspect the asynchronous result and distinguish simulation from production fault injection.

A practical reading path

A new user can begin with the walkthrough, then use the guides to understand a specific workflow. The pre-provision guide covers a virtual baseline and multi-cloud strategy comparison; the RL guide covers an environment and policy-evaluation loop; the chaos guide covers failure hypotheses inside a simulation. After trying an example, consult provider coverage and fidelity to understand which services and behaviors are represented, then review accuracy evidence before relying on modeled estimates.

Use documentation with the API reference

The narrative guides explain why and when to use an endpoint, while the API reference is the source for current request and response schemas. Keep credentials in your local environment rather than copying them into shared examples. Start with a virtual simulation, inspect the returned state and metrics, and make one change at a time so the modeled outcome remains interpretable.