Use the CLI - CloudEval AI Docs

This is the fastest path if you want CloudEval in a terminal, CI job, or internal automation.

CloudEval terminal UI showing project-aware chat, reasoning progress, project context, model, mode, input controls, and credit status

When this path fits

Use the CLI when you want to:

Create or inspect projects without living in the browser
Run reports and capture machine-readable output
Ask one grounded question in a script
Open exact app pages for projects, reports, or billing views

If you mainly need visual diagram review or sharing, the web app is still the better primary surface.

CloudEval command line panel with install command, start commands, and agent handoff actions

Use the CLI in six steps

Install the CLI

Install the npm package on Node.js 20+ machines:

npm install -g @ganakailabs/cloudeval-cli

Or install the standalone release binary:

curl -fsSL https://cli.cloudeval.ai/install.sh | bash

On Windows or Linux with PowerShell 7+:

irm https://cli.cloudeval.ai/install.ps1 | iex

The bash installer may offer to install shell tab completions for your login shell (bash, zsh, or fish). To skip that step, run CLOUDEVAL_INSTALL_COMPLETION=0 curl -fsSL https://cli.cloudeval.ai/install.sh | bash.CloudEval ships these command names:

cloudeval
cloud
eva

Verify the install:

cloudeval --version
eva --version

Configure a profile

Store local defaults for the CloudEval service, app, default project, and model. Profiles are useful for agents, CI jobs, and machines that work with more than one environment.

cloudeval setup \
  --non-interactive \
  --profile quickstart \
  --model gpt-5-nano \
  --format json

Check the profile and local environment:

cloudeval config show --profile quickstart --format json
cloudeval status --profile quickstart --format json
cloudeval doctor --profile quickstart --format json

Enable shell completion

Print the completion script for your shell:

cloudeval completion bash
cloudeval completion zsh
cloudeval completion fish
cloudeval completion powershell

Install completion to a standard per-user path:

cloudeval completion install --shell bash
cloudeval completion install --shell zsh
cloudeval completion install --shell fish
cloudeval completion install --shell powershell

Remove installed completion:

cloudeval completion uninstall --shell bash
cloudeval completion uninstall --shell zsh
cloudeval completion uninstall --shell fish
cloudeval completion uninstall --shell powershell

Choose the login mode based on where the CLI is running.

Workstation
Headless or SSH

Use normal browser login when you are running the CLI on your own laptop or desktop and the terminal can open a browser on the same machine.

cloudeval login
cloudeval auth status

Use headless login when the CLI is running somewhere that cannot open a usable browser window, such as an SSH session, remote server, container, or CI-like terminal.The command prints a browser URL and device code. Open that URL in any browser, approve the session, then return to the terminal.

cloudeval login --headless
cloudeval auth status

Headless login is not a different account type. It is only a login flow for terminals that cannot complete the normal browser handoff.

Create a project

If your source of truth is Bicep, compile it to ARM JSON first. The CLI’s strongest IaC import path today is ARM JSON.

Local ARM JSON
GitHub URL
IaC workspace
Cloud sync

Download an Azure Quickstart template, then create a project from the local file:

curl -fsSL \
  https://raw.githubusercontent.com/Azure/azure-quickstart-templates/master/quickstarts/microsoft.compute/1vm-2nics-2subnets-1vnet/azuredeploy.json \
  -o ./azuredeploy.json

cloudeval projects create \
  --template-file ./azuredeploy.json \
  --name "Azure quickstart import" \
  --provider azure \
  --profile quickstart \
  --format json \
  --output ./cloudeval-project.json

Create a project directly from the Azure Quickstart repository URL:

cloudeval projects create \
  --template-url https://github.com/Azure/azure-quickstart-templates/blob/master/quickstarts/microsoft.compute/1vm-2nics-2subnets-1vnet/azuredeploy.json \
  --name "Azure quickstart import" \
  --provider azure \
  --profile quickstart \
  --format json \
  --output ./cloudeval-project.json

Create a project from a folder that contains the visualization source, linked ARM templates, parameters, and optional .cloudeval/config.yaml:

cloudeval projects create \
  --workspace-dir ./infra \
  --workspace-entry azuredeploy.json \
  --name "Nested ARM workspace" \
  --provider azure \
  --profile quickstart \
  --format json \
  --output ./cloudeval-project.json

If .cloudeval/config.yaml is missing, the CLI creates one using the selected entry and detected parameters file.

Create a Cloud sync project from a least-privilege Azure service principal and one or more resource-group scopes:

cloudeval projects create \
  --cloud-sync \
  --azure-tenant-id "$AZURE_TENANT_ID" \
  --azure-client-id "$AZURE_CLIENT_ID" \
  --azure-client-secret "$AZURE_CLIENT_SECRET" \
  --azure-subscription-id "$AZURE_SUBSCRIPTION_ID" \
  --resource-group rg-app \
  --resource-group rg-network \
  --name "Production Cloud sync" \
  --profile quickstart \
  --format json \
  --output ./cloudeval-project.json

The JSON response includes the created project and connection metadata. Capture both IDs for follow-up commands:

PROJECT_ID=$(jq -r '.data.project.id' ./cloudeval-project.json)
CONNECTION_ID=$(jq -r '.data.connection.id' ./cloudeval-project.json)

cloudeval projects get "$PROJECT_ID" --profile quickstart --format json
cloudeval connections get "$CONNECTION_ID" --profile quickstart --format json

cloudeval projects create supports four creation sources: --template-file, --template-url, --workspace-dir, and --cloud-sync. Use exactly one source per command.

Run reports

cloudeval reports run \
  --project "$PROJECT_ID" \
  --type all \
  --wait \
  --profile quickstart \
  --format json

Fetch a human-readable cost summary:

cloudeval reports cost \
  --project "$PROJECT_ID" \
  --profile quickstart \
  --format markdown

Fetch a human-readable architecture summary:

cloudeval reports waf \
  --project "$PROJECT_ID" \
  --profile quickstart \
  --format markdown

Download saved report payloads for another tool or CI artifact:

cloudeval reports download \
  --project "$PROJECT_ID" \
  --type all \
  --view parsed \
  --output ./reports \
  --profile quickstart \
  --format json \
  --non-interactive

For nested ARM workspace projects, these report commands use the resolved stack generated from .cloudeval/config.yaml. For Cloud sync projects, they use the latest synced Azure snapshot.

Open the project or ask a question

Open the diagram:

cloudeval open project "$PROJECT_ID" \
  --view preview \
  --layout architecture \
  --print-url \
  --no-open

Ask one grounded question:

cloudeval ask "Summarize the top cost and architecture risks in this project" \
  --project "$PROJECT_ID" \
  --profile quickstart \
  --format json \
  --non-interactive

Use ask when you want one answer back. Use chat when you want an ongoing interactive session.

CLI examples for workspace and Cloud sync projects

Nested ARM workspace

cloudeval projects create \
  --workspace-dir ./infra \
  --workspace-entry azuredeploy.json \
  --name "Nested ARM workspace" \
  --provider azure \
  --profile quickstart \
  --format json \
  --output ./cloudeval-project.json

PROJECT_ID=$(jq -r '.data.project.id' ./cloudeval-project.json)

cloudeval projects get "$PROJECT_ID" \
  --profile quickstart \
  --format json

cloudeval projects graph "$PROJECT_ID" \
  --profile quickstart \
  --format json \
  --non-interactive

cloudeval reports run \
  --project "$PROJECT_ID" \
  --type all \
  --wait \
  --profile quickstart \
  --format json

cloudeval reports download \
  --project "$PROJECT_ID" \
  --type all \
  --view parsed \
  --output ./cloudeval-reports \
  --profile quickstart \
  --format json \
  --non-interactive

You can still validate a single local ARM template directly before importing the workspace:

cloudeval validate template \
  --template-file ./azuredeploy.json \
  --parameters-file ./azuredeploy.parameters.json \
  --details \
  --wait \
  --profile quickstart \
  --format json \
  --non-interactive

Cloud sync project

cloudeval projects create \
  --cloud-sync \
  --azure-tenant-id "$AZURE_TENANT_ID" \
  --azure-client-id "$AZURE_CLIENT_ID" \
  --azure-client-secret "$AZURE_CLIENT_SECRET" \
  --azure-subscription-id "$AZURE_SUBSCRIPTION_ID" \
  --resource-group rg-app \
  --name "Production Cloud sync" \
  --profile quickstart \
  --format json \
  --output ./cloudeval-project.json

PROJECT_ID=$(jq -r '.data.project.id' ./cloudeval-project.json)

cloudeval projects get "$PROJECT_ID" \
  --profile quickstart \
  --format json

cloudeval projects graph sync-runs "$PROJECT_ID" \
  --profile quickstart \
  --format json \
  --non-interactive

cloudeval reports run \
  --project "$PROJECT_ID" \
  --type all \
  --wait \
  --profile quickstart \
  --format json

cloudeval reports cost \
  --project "$PROJECT_ID" \
  --profile quickstart \
  --format markdown

cloudeval reports waf \
  --project "$PROJECT_ID" \
  --profile quickstart \
  --format markdown

Inspect models and sessions

List CloudEval-supported models and set a default:

cloudeval models list --profile quickstart --format json
cloudeval models default set gpt-5-nano --profile quickstart --format json

After an ask run, inspect local session history:

cloudeval sessions list --profile quickstart --format json
cloudeval sessions get <thread-id> --profile quickstart --format json

You can also search, rename, and resume recent work:

cloudeval sessions search "cost risk" --profile quickstart --format json
cloudeval sessions rename <thread-id> "Cost and architecture review" --profile quickstart --format json
cloudeval chat --resume "Cost and architecture review" --profile quickstart

For repeatable one-shot follow-ups, reuse a thread ID:

cloudeval ask "What changed since the previous report?" \
  --thread <thread-id> \
  --project "$PROJECT_ID" \
  --profile quickstart \
  --format json \
  --non-interactive

Run CloudEval as an MCP server

If your agent framework already supports MCP, you can expose CloudEval as a local tool server instead of calling one CLI command at a time.

cloudeval mcp status --format json
cloudeval doctor --mcp --format json
codex mcp add cloudeval -- cloudeval mcp serve

CloudEval can generate setup instructions for common MCP clients:

cloudeval mcp setup codex --dry-run
cloudeval mcp setup claude --dry-run
cloudeval mcp setup cursor --dry-run
cloudeval mcp setup generic --dry-run --toolset readonly --format json

Use a focused toolset when an agent only needs part of the CloudEval surface:

cloudeval mcp serve --toolset readonly
cloudeval mcp serve --toolset reports
cloudeval mcp serve --toolset billing

For custom MCP clients, the equivalent JSON shape is:

{
  "mcpServers": {
    "cloudeval": {
      "command": "cloudeval",
      "args": ["mcp", "serve"]
    }
  }
}

For Ollama-powered agents, add this CloudEval MCP entry to the MCP-capable agent host that Ollama launches. The CloudEval server still runs over standard stdio MCP. Use stored cloudeval login auth, stored cloudeval login --headless auth, or --machine. Run login before starting mcp serve; stdin is reserved for MCP protocol messages. MCP clients that support richer discovery can also see CloudEval resources and prompt templates for capabilities, projects, billing summaries, report review, cost review, architecture review, and billing review.

Safe defaults for automation

Run cloudeval capabilities --format json before relying on command names in automation.
Run cloudeval doctor --format json before relying on a local install in automation.
Run cloudeval doctor --mcp --format json before relying on MCP setup.
Prefer --format json for scripts.
Prefer --non-interactive in CI or agent workflows.
Prefer --profile <name> when multiple agents or environments share one machine.
Prefer --print-url --no-open when a command can generate a CloudEval app link.
Use cloudeval login --headless when you are not on a workstation with a usable browser.
Use --machine only when service-principal machine authentication is configured.

Next step

Use CLI overview for the command model, CLI command reference for the current surface area, or Automate evaluations with the CLI for repeatable workflows.

​When this path fits

​Use the CLI in six steps

​CLI examples for workspace and Cloud sync projects

​Nested ARM workspace

​Cloud sync project

​Inspect models and sessions

​Run CloudEval as an MCP server

​Safe defaults for automation

​Next step

When this path fits

Use the CLI in six steps

CLI examples for workspace and Cloud sync projects

Nested ARM workspace

Cloud sync project

Inspect models and sessions

Run CloudEval as an MCP server

Safe defaults for automation

Next step