`json
{
"actor": {
"manifest_path": "/path/to/chat-proxy-actor/manifest.toml"
},
"config": {
"model_config": {
"model": "claude-sonnet-4-20250514",
"provider": "anthropic"
},
"temperature": 1.0,
"max_tokens": 8192,
"system_prompt": "You are a helpful assistant...",
"title": "My Chat Session",
"mcp_servers": [
{
"actor_id": null,
"stdio": {
"command": "/path/to/mcp-server",
"args": ["--arg1", "value1"]
},
"tools": null
}
]
}
}
`

$3

- actor.manifest_path - Path to the Theater actor manifest file
- config.model_config.model - The model to use (e.g., "claude-sonnet-4-20250514")
- config.model_config.provider - The provider (e.g., "anthropic", "openai", "google")

$3

- config.temperature - Sampling temperature (0-2, default varies by model)
- config.max_tokens - Maximum tokens in response
- config.system_prompt - System prompt to set context
- config.title - Display title for the chat session
- config.mcp_servers - Array of MCP server configurations for tool access

Examples

$3

`json
{
"actor": {
"manifest_path": "/path/to/chat-proxy-actor/manifest.toml"
},
"config": {
"model_config": {
"model": "claude-sonnet-4-20250514",
"provider": "anthropic"
},
"title": "General Chat",
"system_prompt": "You are a helpful assistant.",
"mcp_servers": []
}
}
`

$3

`json
{
"actor": {
"manifest_path": "/path/to/chat-proxy-actor/manifest.toml"
},
"config": {
"model_config": {
"model": "claude-sonnet-4-20250514",
"provider": "anthropic"
},
"title": "Programming Assistant",
"system_prompt": "You are a pair programming assistant with filesystem access.",
"mcp_servers": [
{
"stdio": {
"command": "/path/to/fs-mcp-server",
"args": ["--allowed-dirs", "/path/to/project"]
}
}
]
}
}
`

$3

`bash
theater-chat --config config.json --message "Hello! Can you help me review this code?"
`

Interactive Mode

`
┌─ 🎭 Programming Assistant ─────────────────────┐
│ claude-sonnet-4-20250514 │
│ Status: Ready │
└───────────────────────────────────────────────┘

💡 Shortcuts: Ctrl+C (exit), Ctrl+L (clear), Ctrl+T (tool display), Ctrl+H (toggle help)

ℹ️ system: Type your questions or commands.

👤 You: help me understand this codebase

🤖 Assistant: I'd be happy to help you understand the codebase! Let me take a look at the project structure...

🔧 Listing directory

💬 _
`

Architecture

This tool provides a generic inline chat interface for any Theater actor:

- Ink Framework - React-based CLI components for rich terminal UIs
- WebSocket Client - Direct connection to Theater server
- JSON Configuration - Flexible chat-state actor configuration
- Message Compatibility - Full compatibility with Theater chat protocol
- Tool Integration - Support for MCP servers and tool calls
- StartChat Protocol - Automated initialization and setup workflow

$3

theater-chat now follows an enhanced startup sequence for better automation:

1. 🚀 Start Domain Actor - Spawn the proxy actor with configuration
2. 📋 Get Chat State Actor ID - Retrieve the underlying chat actor
3. 🤖 Start Chat Automation - Trigger domain-specific setup (StartChat)
4. 💬 Begin Chatting - Ready for user interaction

This flow allows proxy actors to perform automation tasks (like repository analysis, file indexing, or context setup) before the chat begins, providing users with real-time feedback during initialization.

Development

`bash


Install dependencies

- Enter - Send message
- Ctrl+C - Exit application
- Ctrl+L - Clear message history
- Ctrl+T - Toggle tool call display (hidden/minimal/full)
- Ctrl+H - Toggle help text
- Type exit or quit - Exit application

Configuration Examples

Check out example-config.json for a complete example configuration.

$3

Different models you can use in the config.model_config section:

`json
// Claude Sonnet 4
"model_config": {
"model": "claude-sonnet-4-20250514",
"provider": "anthropic"
}

// GPT-4
"model_config": {
"model": "gpt-4",
"provider": "openai"
}

// Gemini Pro
"model_config": {
"model": "gemini-1.5-pro",
"provider": "google"
}
`

$3

Add tool capabilities via MCP servers:

`json
"mcp_servers": [
{
"stdio": {
"command": "/path/to/filesystem-server",
"args": ["--allowed-dirs", "/project"]
}
},
{
"stdio": {
"command": "/path/to/web-search-server",
"args": ["--api-key", "your-key"]
}
}
]
``

Next Steps

1. ✅ Basic configuration loading
2. ✅ Generic chat interface
3. ✅ Tool call display options
4. Add configuration validation and better error messages
5. Add configuration templates/presets
6. Package for easy distribution
7. Add configuration wizard