IndusAGI TypeScript

TypeScript agent framework with tools, memory, and multi-agent orchestration.

Install

npm install indusagi-ts

Quick Start

import { Agent, AgentConfig, registry, registerDefaultTools } from "indusagi-ts";

const config = AgentConfig.fromEnv();
const agent = new Agent({
  name: "Helper",
  role: "Helpful assistant",
  config
});

registerDefaultTools();
agent.context = registry.context;

const response = await agent.processWithTools({
  userInput: "What time is it and what is 25 * 4?",
  tools: registry.schemas,
  toolExecutor: registry
});

console.log(response);

Multi-Agent Agency

import { createImprovedAgency, registry } from "indusagi-ts";

const agency = createImprovedAgency({
  model: "gpt-4o",
  provider: "openai"
});

const result = await agency.process({
  userInput: "Plan and implement a small CLI tool",
  useTools: true,
  tools: agency.tools,
  toolExecutor: registry
});

console.log(result.response);

CLI

indusagi run "What is 25 * 4?"
indusagi interactive
indusagi list-tools
indusagi test-connection
indusagi agency-demo
indusagi tui

Configuration

Set provider API keys in your environment:

• OPENAI_API_KEY
• ANTHROPIC_API_KEY
• GROQ_API_KEY
• OLLAMA_API_KEY
• GEMINI_API_KEY / GOOGLE_API_KEY
• MISTRAL_API_KEY

Optional:

• LLM_PROVIDER (openai, anthropic, groq, ollama, google, mistral)
• OPENAI_BASE_URL, ANTHROPIC_BASE_URL, GROQ_BASE_URL, OLLAMA_BASE_URL, GOOGLE_GENAI_USE_VERTEXAI
• MAX_TOKENS, TEMPERATURE, MAX_TURNS

TUI

The TUI uses a terminal UI powered by blessed. Run:

indusagi tui

Use key bindings shown in the footer (Ctrl+N, Ctrl+S, Ctrl+M, Ctrl+T, Ctrl+P, Ctrl+Q).

Development

npm run build
npm run typecheck
npm run test

MCP Integration

MCP tooling is supported via optional configuration in code. Load tools from a mcp.json file with:

import { loadMcpJson, McpToolRouter } from "indusagi-ts";

const config = loadMcpJson("./mcp.json");
const router = new McpToolRouter({ config });
const tools = await router.listTools();