Quick Start

This guide walks you through the 3-step onboarding process to get your first tool running. Complete this guide in 2 minutes.

Prerequisites

Install 2LY and ensure it's running at http://localhost:8888

Step 1: Create Your Workspace

When you first access http://localhost:8888, you'll be guided through workspace creation:

1. Enter your admin account details
2. Create your workspace

Your workspace is where all your tools, agents, and runtimes are managed. Think of it as your AI tool control center.

Step 2: Follow the Onboarding

2LY provides a 3-step guided onboarding that walks you through the core concepts:

1. Understanding Runtimes - See how distributed execution environments work
2. Connecting MCP Servers - Learn how to add tool sources
3. Creating Toolsets - Discover how to curate tools for your agents

The onboarding is interactive and takes about 2 minutes. Don't skip it—it covers everything you need to know.

Step 3: Connect an MCP Server

2LY comes with a selection of popular MCP servers ready to install:

Recommended for first-time users:

• Filesystem - Let agents read/write files
• GitHub - Connect to repositories
• Weather - Get weather data

To connect a server:

1. Click Browse MCP Servers in the Sources page
2. Select a server from the list (start with Filesystem for simplicity)
3. Configure any required parameters (e.g., allowed directories)
4. Click Test Server to verify it works
5. Click Save to activate it

You can add more MCP servers later from the MCP Server Registry.

Step 4: Test Your Tools

Before connecting your agent, verify your tools work:

1. Navigate to the Tool Tester page
2. Select a tool from your connected MCP server
3. Provide test inputs
4. Click Test and verify the output

This ensures everything is configured correctly before integrating with your agent.

Step 5: Connect Your Agent

Now your toolset is ready to be consumed by agents:

1. Navigate to the Toolsets page
2. Click Connect on your toolset
3. Choose your agent framework:
   • LangChain - Copy the Python code snippet
   • JSON - Use the raw MCP configuration
   • Other frameworks - Use the MCP endpoint URL
4. Follow the framework-specific integration instructions

Example: LangChain Integration

from twoly import MCPAdapter

# Connect to 2LY
adapter = MCPAdapter(
    endpoint="http://localhost:3000/mcp",
    workspace_id="your-workspace-id",
    toolset_id="your-toolset-id"
)

# Use with LangChain
from langchain.agents import create_openai_tools_agent

agent = create_openai_tools_agent(
    llm=llm,
    tools=adapter.get_tools()
)

See the LangChain Integration Guide for complete instructions.

Verification

You should now have:

• ✅ Workspace created with admin account
• ✅ At least one MCP server connected
• ✅ Tools tested and verified
• ✅ Agent integration code ready

What Happens Next?

Now that you've completed the quick start, here's what to explore:

Learn More About Tools

• MCP Servers - Explore available MCP servers
• Tool Capabilities - Understand what tools can do
• API & REST - Connect to REST APIs
• Coded Functions - Write custom tool code

Agent Integrations

• LangChain - Complete LangChain setup guide
• LangFlow - Visual workflow builder
• N8N - Automation platform integration

Deep Dives

• Architecture - How 2LY works under the hood
• Runtime Deployment - Deploy runtimes anywhere
• Security - Secure your 2LY installation

Production Deployment

Ready to deploy 2LY for production use?

• Scale runtimes across multiple environments
• Configure monitoring and observability
• Set up proper authentication and access controls
• Deploy to cloud platforms (AWS, GCP, Azure)

See the Deployment Guide for production setup instructions.

Troubleshooting

Can't access localhost:8888?

# Check if services are running
docker compose ps

# View logs
docker compose logs -f frontend

# Restart services if needed
docker compose restart

MCP test fails?

• Check directory permissions for filesystem-based servers
• Try /tmp instead of Desktop for testing
• Verify the MCP server configuration in the UI
• Check runtime logs: docker compose logs tool-runtime

Tools not appearing in agent?

• Ensure your agent is properly configured with the MCP endpoint
• Check that the toolset includes at least one tool
• Verify the runtime is connected in the dashboard