How to Connect Custom Coder Agent

This comprehensive guide shows you how to connect and use your custom coder agent from UBOS Agentspace directly in VS Code using the Cline extension. This integration allows you to leverage your custom AI agent for coding assistance, debugging, and development tasks directly in your IDE.

---

Overview

The Custom Coder Agent integration enables you to:

• Access your UBOS Agentspace custom agents directly from VS Code
• Use specialized coding agents with custom knowledge and capabilities
• Leverage your agent's custom actions and topics for development tasks
• Maintain context and conversation history across coding sessions

Prerequisites

Before you begin, ensure you have:

• An active UBOS Agentspace account
• A custom coder agent created in your Agentspace (see How to Create an Agent)
• VS Code installed on your system
• The Cline extension installed in VS Code

---

Step 1: Access Your Custom Agent in Agentspace

1. Sign in to UBOS Agentspace
   First, sign in to your UBOS Agentspace platform using your credentials.

2. Navigate to My Agents
   In your UBOS Agentspace dashboard, click on My Agents from the left sidebar to view all your available agents.

3. Locate Your Coder Agent
   Find your custom coder agent in the agents list. You can identify it by the name and description you provided during creation.

4. Access Agent Advanced Settings
   Click on your agent to open its details page, then navigate to the Advanced tab in the left sidebar.

---

Step 2: Configure Custom LLM Model

In the Advanced settings, you'll see the Custom LLM Model Management section with important connection details:

Model Configuration Details

The system will display your agent's connection information:

• Model Running Status: Indicates if your model is active
• LiteLLM URL: The endpoint URL for your agent (e.g., https://litellm-1819-68a7d218026d9811000012e.agentspaceuat.ubos.tech/)
• Token: Your authentication token (displayed as sk-x)
• Model Name: Your agent's identifier (e.g., qwen3-coder-dt)

Important Notes

• Keep your token secure and never share it publicly
• The LiteLLM URL is unique to your agent and workspace
• Your model must be in "Running" status to accept connections

---

Step 3: Install and Configure Cline Extension

Install Cline Extension

1. Open VS Code Extensions
   Launch VS Code and click on the Extensions icon in the left sidebar (or press Ctrl+Shift+X).

2. Search for Cline
   In the extensions marketplace, search for "Cline" and install the extension.

3. Verify Installation
   After installation, you should see the Cline icon in your VS Code sidebar.

Configure Cline with Your Custom Agent

1. Open Cline Settings
   Click on the Cline extension icon, then click on the settings/configuration button.

2. Select API Provider
   Choose "OpenAI Compatible API Key" as your provider type.

3. Enter Configuration Details
   Fill in the following information from your Agentspace Advanced settings:

   Field	Value	Example
   Base URL	Your LiteLLM URL from Agentspace	https://litellm-1819-68a7d218026d9811000012e.agentspaceuat.ubos.tech/
   API Key	Your token from Agentspace	sk-x (use your actual token)
   Model ID	Your model name	qwen3-coder-dt
   Context Window Size	Set according to your model's capabilities	65000 (for qwen3-coder-dt)

4. Configure Advanced Options (Optional)
   You can adjust additional settings such as:

   • Temperature: Controls response creativity (0.0 - 1.0)
   • Max Output Tokens: Maximum response length

---

Step 4: Test Your Connection

Verify Connection

1. Start a New Chat
   Open Cline and start a new conversation to test the connection.

2. Send a Test Message
   Try a coding-related query like the one shown in the screenshot:

   Can you create me a good tech company landing page?

3. Verify Response
   Your custom agent should respond according to its configured personality and capabilities, potentially offering to create HTML, CSS, and JavaScript files for a modern landing page.

Test Agent Capabilities

Test various aspects of your custom agent with practical requests:

• Web Development: Ask it to create landing pages, components, or web applications
• Code Generation: Request specific functions, classes, or complete file structures
• Code Review: Have it analyze existing code in your workspace
• Project Setup: Ask for help setting up new projects or frameworks
• Debugging: Request help with error messages or issues

Example Test Queries

Try these example queries to test different capabilities:

Can you help me create a responsive navigation bar for my website?

Build a REST API with authentication using Node.js and Express

Create a React component for a product card with hover effects

---

Troubleshooting

Common Issues and Solutions

Issue	Possible Cause	Solution
Connection & Auth Failed	Incorrect URL or token	Verify your LiteLLM URL and token in Agentspace Advanced settings
Model Not Responding	Agent is stopped	Check that your agent status shows "Running" in Agentspace
Slow Responses	Network or model load	Check your internet connection and agent status
Unexpected Responses	Wrong model configuration	Verify the model ID matches your agent name

Getting Help

If you encounter issues:

1. Check Agent Status: Ensure your agent is running in Agentspace
2. Verify Configuration: Double-check all connection details
3. Review Logs: Check VS Code's output panel for error messages

For more information on customizing your agent's capabilities, see:

• How to Create a Topic
• How to Create an Action
• Agent Concepts

---

Pro Tip

Create different agents for different types of coding tasks (frontend, backend, DevOps, etc.) and switch between them in Cline based on your current work context.