CLI Reference
The curate CLI provides command-line access to the Curate-Me AI Gateway platform. Manage agents, runners, costs, workflows, API keys, and more from your terminal.
Installation
npm install -g @curate-me/cli
# or
npx @curate-me/cliRequires Node.js 18 or later.
Authentication
# Log in to the platform (opens browser for OAuth)
curate login
# Check current authentication status
curate whoami
# Log out
curate logoutAfter logging in, your credentials are stored locally and used for all subsequent commands.
Configuration
# Initialize a new project with .curateme.json config
curate init
# View current configuration
curate config
# Set configuration values
curate config set apiUrl http://localhost:8001
curate config set orgId org_xxxGateway Commands
API Keys
# List API keys
curate keys list
# Create a new API key
curate keys create --name "production" --scopes chat,admin
# Revoke an API key
curate keys revoke <key-id>Cost Tracking
# View cost summary for current period
curate costs
# View costs for a specific period
curate costs --period=week
curate costs --period=month
# View costs in JSON format
curate costs --jsonAgent Commands
# List all agents
curate agents list
curate agents list --status active
curate agents list --json
# Get agent details
curate agents get <agent-id>
curate agents get <agent-id> --json
# Run an agent
curate agents run <agent-id> --input '{"query": "Hello"}'
curate agents run <agent-id> --file input.json
curate agents run <agent-id> --stream # Stream output in real-time
# Run agent tests
curate agents test <agent-id>
# View agent logs
curate agents logs <agent-id>
curate agents logs <agent-id> --follow # Stream live logs
curate agents logs <agent-id> --lines 100
# View version history
curate agents versions <agent-name>
# Rollback to a previous version
curate agents rollback <agent-name> <version>
curate agents rollback <agent-name> 3 --yes # Skip confirmation
# Configure model fallback chain
curate agents fallback <agent-name> --list
curate agents fallback <agent-name> --apply-defaultRunner Commands
Manage the full lifecycle of managed OpenClaw sandbox runners.
Core Lifecycle
# List runners
curate runners list
curate runners list --state running
curate runners list --json
# Create a new runner
curate runners create
curate runners create --profile locked # Most restrictive
curate runners create --profile web_automation # Browser automation
curate runners create --profile full_vm_tools # Full access
# Get runner details
curate runners get <runner-id>
curate runners get <runner-id> --json
# Start a session on a runner
curate runners start <runner-id>
# Stop a session
curate runners stop <runner-id>
# Execute a command in a runner
curate runners exec <runner-id> "echo hello"
curate runners exec <runner-id> "npm test" --timeout 60
# View audit events for a runner
curate runners events <runner-id>
curate runners events <runner-id> --limit 20
# Stream real-time output from a runner
curate runners stream <runner-id>
# Terminate a runner (stops all sessions, releases resources)
curate runners terminate <runner-id>
# Check runner subsystem health
curate runners healthDesktop Streaming
# Capture a desktop screenshot
curate runners desktop screenshot <runner-id>
# Get a stream token for browser-based desktop viewing
curate runners desktop stream-token <runner-id>Your Machines
# List your connected machines
curate machines list
# Connect a new machine (generates token + prints install steps)
curate machines connect
# Check detailed status of a connected machine
curate machines status <machine-id>
# Remove a machine from your organization
curate machines remove <machine-id>Agent Communication
# Chat with a deployed agent interactively
curate runners agent chat <agent-id> "Fix the failing test"
# Check agent status
curate runners agent status <agent-id>
# Send a file to an agent
curate runners agent send-file <agent-id> ./report.pdf
# Dispatch a task and optionally watch output
curate runners agent task <agent-id> "Refactor auth module"
curate runners agent task <agent-id> "Refactor auth module" --watch
# List tasks for an agent
curate runners agent tasks <agent-id>
# List files in an agent workspace
curate runners agent files <agent-id>
# Take a screenshot of the agent desktop
curate runners agent screenshot <agent-id>
# Listen for real-time events from an agent
curate runners agent listen <agent-id>Workflow Commands
# List workflows
curate workflows list
curate workflows list --status active
# Get workflow details
curate workflows get <workflow-id>
# Run a workflow
curate workflows run <workflow-id> --input '{"key": "value"}'
curate workflows run <workflow-id> --stream
# View workflow execution history
curate workflows executions <workflow-id>Evaluation Commands
# List evaluation runs
curate eval list
# Run an evaluation
curate eval run --config eval-config.json
# Get evaluation results
curate eval get <eval-id>Organization Commands
# View current organization
curate org
# Switch organization context
curate org switch <org-id>
# List organization members
curate org membersDeployment Commands
# Deploy to an environment
curate deploy
curate deploy --environment staging
curate deploy --environment productionLog Commands
# View platform logs
curate logs
curate logs --follow
curate logs --level error
curate logs --source gatewayGlobal Options
These options are available on all commands:
| Option | Description |
|---|---|
--json | Output in JSON format (machine-readable) |
--verbose | Enable verbose output |
--quiet | Suppress non-essential output |
--no-color | Disable colorful output |
-v, --version | Show CLI version |
-h, --help | Show help |
Configuration File
The curate init command creates a .curateme.json file in your project root:
{
"apiUrl": "https://api.curate-me.ai",
"orgId": "org_xxx",
"outputFormat": "table"
}Environment variables override config file values:
| Variable | Description |
|---|---|
CURATE_ME_API_KEY | API key for authentication |
CURATE_ME_ORG_ID | Organization ID |
CURATE_ME_API_URL | API base URL |