The frontend coding agent for production codebases

Agent Interface

This package offers both interface definitions and base-functionality to integrate agents with stagewise.

You can find more information on how to use this interface in the following guide: Build custom Agent Integrations

Features

• Availability Management: Control agent availability with error handling
• State Management: Track agent state (idle, processing, etc.)
• Messaging: Handle user messages and stream agent responses
• Chat Capability: Full chat history support with multi-chat management (NEW!)

Chat Capability

The new chat capability provides comprehensive chat functionality:

Key Features

• Multi-Chat Support: Create and manage multiple chat sessions
• Message History: Full conversation history with user, assistant, and tool messages
• Streaming Updates: Stream message parts in parallel for responsive UI
• Tool Integration: Tools are now integrated within chat messages
• State Management: Only one chat can be active at a time

Message Types

Aligned with Vercel AI SDK:

• UserMessage: Text, image, and file content with browser metadata
• AssistantMessage: Text, files, reasoning, tool calls, and tool results
• ToolMessage: Tool execution results

Usage Example

import { createAgentServer } from '@stagewise/agent-interface';

const { agent } = await createAgentServer({ port: 3000 });

// Enable chat support
agent.chat.setChatSupport(true);

// Create a chat
const chatId = await agent.chat.createChat('My Chat');

// Listen for updates
agent.chat.addChatUpdateListener((update) => {
  console.log('Chat update:', update);
});

// Handle messages
agent.messaging.addUserMessageListener(async (msg) => {
  // Stream assistant response
  agent.chat.streamMessagePart('msg-1', 0, {
    content: { type: 'text', text: 'Hello!' },
    updateType: 'create',
  });
});

See examples/chat-usage.ts for a complete example.