Ollama Agent

An AI coding agent VS Code extension powered by Ollama running locally. This extension provides an intelligent coding assistant that can read, write, and modify files in your workspace, execute shell commands, fetch web content, and integrate with MCP (Model Context Protocol) servers for extended functionality.

Features

• 🤖 AI-Powered Coding Assistant: Leverages local LLMs via Ollama for code generation, refactoring, and analysis
• 💬 Interactive Chat Interface: Sidebar chat panel with conversation history and context management
• 🛠️ File Operations: Read, create, edit, and delete files in your workspace
• ⚡ Command Execution: Run shell commands directly from the agent (PowerShell on Windows, bash/zsh on macOS/Linux)
• 🌐 Web Fetching: Retrieve and analyze web content
• 🔌 MCP Server Support: Extend capabilities with Model Context Protocol servers
• 🔄 Streaming Responses: Real-time response streaming with stop capability
• 📊 Context Window Tracking: Monitor token usage and remaining context

Requirements

• VS Code 1.85.0 or higher
• Ollama installed and running locally
• At least one model pulled in Ollama (e.g., qwen2.5-coder:32b)

Installation

1. Clone this repository:

   git clone <repository-url>
   cd ollama-agent
   

2. Install dependencies:

   npm install
   

3. Compile the extension:

   npm run compile
   

4. Press F5 to open a new Extension Development Host window

Configuration

Configure the extension through VS Code settings (Ctrl+, or Cmd+,):

Example MCP Server Configuration

{
	"ollamaAgent.mcpServers": [
		{
			"name": "filesystem",
			"command": "npx",
			"args": ["-y", "@modelcontextprotocol/server-filesystem", "/path/to/allowed/dir"],
			"enabled": true
		}
	]
}

Usage

Opening the Chat

• Click the Ollama icon in the Activity Bar
• Use the status bar button: $(comment-discussion) Ollama Chat
• Run the command: Ollama Agent: Chat

Running a Task

1. Open the chat panel
2. Select your desired model from the dropdown
3. Type your request (e.g., "Create a React component for a todo list")
4. The agent will:
   • Analyze your request
   • Read existing files if needed
   • Create or modify files
   • Execute commands (build, test, etc.)
   • Provide a summary of actions taken

Available Commands

Agent Capabilities

The agent can perform the following actions:

Supported Models

The extension works with any Ollama model that supports chat completions. Recommended models:

• qwen2.5-coder:32b (default) - Excellent for coding tasks
• codellama:34b - Good for code generation
• deepseek-coder:33b - Strong coding capabilities
• llama3.1:70b - General purpose with good reasoning

Development

Project Structure

ollama-agent/
├── src/
│   ├── extension.ts          # Main extension entry point
│   ├── agent/
│   │   ├── agent.ts          # Agent orchestration
│   │   ├── executor.ts       # Tool execution logic
│   │   ├── llm.ts            # LLM provider (Ollama)
│   │   ├── mcp.ts            # MCP server management
│   │   └── tools.ts          # Tool definitions and prompts
│   ├── utils/
│   │   └── workspace.ts      # Workspace utilities
│   └── test/
│       └── mockProvider.ts   # Test utilities
├── package.json              # Extension manifest
└── tsconfig.json             # TypeScript configuration

Building

# Compile TypeScript
npm run compile

# Watch for changes
npm run watch

# Package for distribution
npm run package

Debugging

1. Open the project in VS Code
2. Set breakpoints in the source code
3. Press F5 to launch the Extension Development Host
4. Use the extension in the new window
5. Debugging output appears in the Debug Console

Platform Support

The extension automatically adapts to your operating system:

Troubleshooting

Ollama Connection Issues

1. Ensure Ollama is running: ollama serve
2. Verify the endpoint URL in settings
3. Check that the model is downloaded: ollama list

Model Not Responding

1. Check the Output panel (View → Output → Ollama Agent)
2. Verify the model name is correct
3. Try a different model if the current one hangs

MCP Server Errors

1. Check the server command is installed and in PATH
2. Verify the arguments are correct
3. Check the Output panel for error messages

License

MIT

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Acknowledgments

• Ollama for making local LLMs accessible
• Model Context Protocol for extending AI capabilities