TrikHub
TrikHub/Docs

Quickstart

Get a working example running in minutes.

The trik create-agent command scaffolds a complete agent project with LangGraph integration and gateway setup pre-configured.

1. Install the CLI

npm install -g @trikhub/cli

2. Scaffold an Agent

trik create-agent ts
cd my-agent
npm install

3. Install a Trik

# Optional Trik to manage triks directly through CLI
trik install @molefas/trikster
 
# Demo trik
trik install @molefas/trik-demo-notes

4. Run

npm run dev

That’s it — you have a working agent with trik integration. The scaffolded project includes a conversation loop, so you can start chatting right away.

Option 2: Add to Existing Project

1. Install Dependencies

npm install -g @trikhub/cli
npm install @trikhub/gateway

The @trikhub/gateway package includes everything you need to load and run triks. You do not need to install @trikhub/manifest separately — it is included as a transitive dependency.

2. Install a Trik

# Optional Trik to manage triks directly through CLI
trik install @molefas/trikster
 
# Demo trik
trik install @molefas/trik-demo-notes

This checks the TrikHub registry, adds the trik to your package.json via a git URL, and registers it in .trikhub/config.json.

3. Integrate with Your Agent

The fastest way to integrate is using the LangChain adapter’s enhance() function, which wraps your LangGraph agent with handoff routing:

import { createReactAgent } from '@langchain/langgraph/prebuilt';
import { ChatOpenAI } from '@langchain/openai';
import { enhance } from '@trikhub/gateway/langchain';
 
const model = new ChatOpenAI({ model: 'gpt-4o-mini' });
 
// enhance() creates the gateway, loads triks, and manages the agent lifecycle
const app = await enhance(null, {
  createAgent: (trikTools) =>
    createReactAgent({
      llm: model,
      tools: [...yourOwnTools, ...trikTools],
    }),
});
 
const response = await app.processMessage('Find me some AI articles');
console.log(response.message);  // The trik's response
console.log(response.source);   // "main", trik ID, or "system"

If you need more control over gateway setup:

import { createReactAgent } from '@langchain/langgraph/prebuilt';
import { ChatOpenAI } from '@langchain/openai';
import { enhance } from '@trikhub/gateway/langchain';
import { TrikGateway } from '@trikhub/gateway';
 
const gateway = new TrikGateway();
await gateway.initialize();
await gateway.loadTriksFromConfig();
 
const model = new ChatOpenAI({ model: 'gpt-4o-mini' });
const app = await enhance(null, {
  gatewayInstance: gateway,
  createAgent: (trikTools) =>
    createReactAgent({
      llm: model,
      tools: [...yourOwnTools, ...trikTools],
    }),
});
 
const response = await app.processMessage('Find me some AI articles');

4. Build a Conversation Loop

The processMessage method handles the full routing lifecycle. Here is a minimal conversation loop:

import * as readline from 'node:readline';
 
const rl = readline.createInterface({ input: process.stdin, output: process.stdout });
 
function ask(prompt: string): Promise<string> {
  return new Promise((resolve) => rl.question(prompt, resolve));
}
 
while (true) {
  const userMessage = await ask('You: ');
  if (userMessage === 'exit') break;
 
  const response = await app.processMessage(userMessage);
  console.log(`[${response.source}]: ${response.message}`);
}

What Happens

  1. User sends a message
  2. The gateway checks if there is an active handoff — if so, the message goes directly to the trik agent
  3. If no handoff is active, the message goes to your main agent
  4. Your main agent sees talk_to_<trik> tools and exposed tools in its tool set
  5. If the agent calls a handoff tool, the gateway starts a handoff session with the trik
  6. The trik handles the conversation autonomously until it transfers back
  7. On transfer-back, a structured session summary is injected into the main agent’s history

Next Steps