chat-agent

chat-agent is the easiest way to create an AI chatbot using TypeScript.

• Simple config: Define your questions and completion logic in a single object.
• LLM-powered: Uses OpenAI or Claude (Anthropic) to extract and validate answers from free-form user input.
• Conversational: Handles multi-question conversations, lets users answer in natural language, and guides them to complete all required info.
• Rapid prototyping: Perfect for forms, lead capture, support, surveys, and more—no need to build a UI or complex logic.

Test Your Own Agent

You can test your own agent by creating a config file (JavaScript or TypeScript) that exports a bot agent instance (e.g., export const bot = chatAgent({...})).

You'll be prompted for your OpenAI API key and can interact with a sample plane ticket booking agent.

Then run:

npx chat-agent --config ./my-agent.js

Example agent file (my-agent.ts)

import { chatAgent } from "chat-agent";

export const bot = chatAgent({
  modelProvider: {
    openai: {
      apiKey: process.env.OPENAI_API_KEY!,
      model: "gpt-4o",
    },
  },
  questions: [
    { id: "name", prompt: "What's your full name?" },
    { id: "from", prompt: "From which city are you departing?" },
    { id: "to", prompt: "Where are you flying to?" },
    { id: "date", prompt: "When do you want to travel?" },
    { id: "roundTrip", prompt: "Is this a round trip or one-way?" },
  ],
  onComplete: async (data) => {
    console.log("Collected data:", data);
    // { name: "...", from: "...", to: "...", date: "...", roundTrip: "..." }
  },
});

More Example Agent Files

Sales Lead Capture Bot (sales-lead-bot.ts)

  questions: [
    { id: "name", prompt: "What's your full name?" },
    { id: "company", prompt: "What company do you represent?" },
    { id: "email", prompt: "What's your business email address?" },
    {
      id: "interest",
      prompt: "What product or service are you interested in?",
    },
    { id: "budget", prompt: "What is your estimated budget?" },
  ],
  onComplete: async (data) => {
    await crm.createLead(data);
    console.log("Sales lead captured:", data);
    // { name: "...", company: "...", email: "...", interest: "...", budget: "..." }
  },

Scheduling Bot (scheduling-bot.ts)

  questions: [
    { id: "name", prompt: "What's your name?" },
    { id: "email", prompt: "What's your email address?" },
    { id: "date", prompt: "What date would you like to schedule?" },
    { id: "time", prompt: "What time works best for you?" },
    { id: "purpose", prompt: "What's the purpose of the meeting?" },
  ],
  onComplete: async (data) => {
    await calendar.schedule(data);
    console.log("Scheduling request:", data);
    // { name: "...", email: "...", date: "...", time: "...", purpose: "..." }
  }

Customer Support Triage Bot (support-triage-bot.ts)

  questions: [
    { id: "name", prompt: "May I have your name?" },
    { id: "email", prompt: "What's your email address?" },
    {
      id: "issueType",
      prompt:
        "What type of issue are you experiencing? (e.g., billing, technical, account)",
    },
    { id: "description", prompt: "Please describe your issue in detail." },
    { id: "urgency", prompt: "How urgent is this issue? (low, medium, high)" },
  ],
  onComplete: async (data) => {
    await zendesk.createTicket(data);
    console.log("Support ticket:", data);
    // { name: "...", email: "...", issueType: "...", description: "...", urgency: "..." }
  }

Data Collection / Survey Bot (survey-bot.ts)

  questions: [
    { id: "age", prompt: "What is your age?" },
    { id: "gender", prompt: "What is your gender?" },
    {
      id: "satisfaction",
      prompt: "How satisfied are you with our service? (1-5)",
    },
    { id: "feedback", prompt: "Any additional feedback or comments?" },
  ],
  onComplete: async (data) => {
    await db.saveSurvey(data);
    console.log("Survey response:", data);
    // { age: "...", gender: "...", satisfaction: "...", feedback: "..." }
  }