Skip to main content
The run command starts Claude Code locally with real-time streaming to the Anyware cloud. This is the default command - running anyware is equivalent to anyware run.

​
Usage

anyware [run] [options] [claude-args...]

​
Options

OptionDescription
-p, --path <path>Working directory (default: current directory)
-m, --model <model>Model to use via OpenRouter (e.g., openai/gpt-4o)
--resume <sessionId>Resume a Claude Code session
--continueContinue the last conversation
--remoteStart in remote mode

​
Examples

​
Basic usage

Start Claude Code in the current directory: 
anyware

​
Specify working directory

anyware --path /path/to/project

​
Use a specific model

anyware --model openai/gpt-4o

​
Resume a previous session

anyware --resume abc123

​
Continue last conversation

anyware --continue

​
Start in remote mode

Start with web dashboard in control: 
anyware --remote

​
Pass Claude Code arguments

Additional arguments are passed through to Claude Code: 
anyware --verbose

​
What Happens

  1. Login check - Verifies you’re authenticated
  2. Session creation - Creates a session on the Anyware server
  3. WebSocket connection - Connects for real-time streaming
  4. Claude Code spawn - Starts Claude Code locally with hooks
  5. Mode loop - Handles switching between local and remote modes

​
Mode Switching

By default, you start in local mode. See Local vs Remote Mode for details on:
  • How input routing works
  • Switching between modes
  • Using the --remote flag

​
Environment

The command sets up Claude Code with:
VariableValue
ANTHROPIC_BASE_URLAnyware LLM proxy URL
ANTHROPIC_AUTH_TOKENYour authentication token
This routes LLM requests through the Anyware proxy for model support and usage tracking.

​
Exit

Exit by:
  • Ctrl+C - Graceful shutdown
  • Exiting Claude Code - Normal exit
The session is ended on the server and the WebSocket connection is closed.

​
See Also