CodeBuddy: Autonomous AI Software Engineer

Not just a copilot—a fully autonomous software engineer that lives in your IDE.

CodeBuddy transforms VS Code into an agentic development environment. Unlike traditional coding assistants that only provide autocomplete or snippets, CodeBuddy is an autonomous agent capable of planning, executing, debugging, and deploying complex software tasks. It understands your entire codebase, runs terminal commands, manages files, and fixes its own errors.

Overview

CodeBuddy transforms developer productivity by acting as a proactive teammate. It goes beyond simple code suggestions to provide deep architectural insights, autonomous refactoring, comprehensive testing strategies, and self-healing workflows. The extension supports both cloud-based and local AI models, enabling developers to choose the right balance of capability, speed, and privacy.

Key Features

Multi-Agent Architecture

CodeBuddy employs specialized agents that collaborate on complex tasks:

• Project Manager: Orchestrates workflow and delegates tasks to specialized agents.
• Architect: Designs scalable systems, selects patterns, and creates architecture decision records.
• Reviewer: Enforces code quality, security standards, and best practices.
• Tester: Writes and validates unit and integration tests.
• Debugger: Investigates root causes and proposes fixes.
• Code Analyzer: Scans codebase for anti-patterns and complexity issues.
• Documentation Writer: Generates comprehensive documentation and API references.
• File Organizer: Refactors directory structures and manages project organization.

Autonomous Self-Healing

When a task fails (e.g., a build error or failed test), CodeBuddy doesn't give up. It follows a rigorous loop:

1. Reads the error output from the terminal.
2. Analyzes the root cause.
3. Applies a fix.
4. Retries the verification.
5. Repeats until success.

Dual Operating Modes

Chat Mode: Traditional question-and-answer interaction for quick queries, code explanations, and getting code snippets without file modifications.

Agent Mode: Autonomous execution with full tool access including file operations, terminal commands, web search, and codebase analysis. Changes can be reviewed before application.

AI Provider Support

CodeBuddy supports nine AI providers, giving you flexibility in model selection:

Provider	Default Model	Capabilities
Gemini	gemini-2.5-pro	Long context, general coding
Anthropic	claude-sonnet-4-5	Complex architecture, refactoring
OpenAI	gpt-4o	Reasoning, planning
DeepSeek	deepseek-chat	Cost-effective coding
Qwen	qwen-max	Strong open-weight performance
Groq	llama-3.1-70b-versatile	Ultra-fast inference
GLM	glm-4	Chinese and English support
XGrok	grok	Alternative reasoning
Local	qwen2.5-coder	Privacy-first, offline capable

Local Model Integration

Run completely offline with local models via Ollama or LM Studio:

• Full Agent mode support with local models.
• Zero API costs for unlimited usage.
• Code never leaves your machine.
• Supports Qwen 2.5 Coder, Llama 3.2, DeepSeek Coder, and other OpenAI-compatible models.

Terminal Integration

Agents can execute shell commands directly:

• Run builds, tests, git operations, and any shell command.
• Real-time output streaming in dedicated terminal channel.
• Safety guards with user confirmation before execution.
• Feedback loop for self-correction based on command output.

Diff Review System

File changes made by the agent are tracked and reviewable:

• Side-by-side diff viewer in VS Code.
• Apply or reject individual changes.
• Recent changes history panel.
• Optional approval mode requiring explicit confirmation before file writes.

Model Context Protocol (MCP) Integration

Extend CodeBuddy's capabilities through the open MCP standard:

• Connect external tools and data sources.
• Add custom MCP servers for specialized functionality.
• Unified protocol for AI-tool integration.
• Supported integrations include PostgreSQL, AWS, Kubernetes, Linear, and more.

Project Rules

Customize agent behavior with project-specific rules:

• Define coding conventions and guidelines in .codebuddy/rules.md.
• Create custom rules via the Settings UI.
• Toggle rules on/off without deletion.
• Rules are appended to the agent's system prompt.

Deep Context Understanding

Intelligent context handling ensures relevant responses:

• Semantic Search: Uses vector embeddings to find relevant code chunks.
• AST Analysis: Understands function definitions and module interactions.
• Token Budget Awareness: Optimizes context usage based on model limits.
• Priority System: Prioritizes @mentioned files, active files, and auto-gathered context.

Installation

Install from your preferred registry:

• VS Code Marketplace
• Open VSX Registry

Alternatively, search for "CodeBuddy" in the VS Code extension manager.

Configuration

Cloud Providers

1. Open VS Code Settings (File > Preferences > Settings).
2. Search for "CodeBuddy".
3. Select your preferred AI provider from the dropdown.
4. Enter your API key in the appropriate field.

Local Models

Configure for Ollama or LM Studio:

{
  "generativeAi.option": "Local",
  "local.baseUrl": "http://localhost:11434/v1",
  "local.model": "qwen2.5-coder"
}

For Docker-based Ollama:

docker compose -f docker-compose.yml up -d
docker exec -it ollama ollama pull qwen2.5-coder

Settings Panel

Access comprehensive settings via the gear icon in the sidebar:

• Account: Profile and account information.
• General: Theme, language, font settings, nickname, streaming preferences.
• Agents: Auto-approve actions, file/terminal permissions, verbose logging.
• MCP: Model Context Protocol server management.
• Conversation: Chat display preferences and history management.
• Models: AI model selection and configuration.
• Context: Workspace indexing, context window size, file inclusion settings.
• Rules and Subagents: Custom rules and specialized agent configuration.
• Privacy: Data privacy settings and clear data options.
• Beta: Experimental features.
• About: Version information and links.

Data Storage

CodeBuddy stores data locally in a .codebuddy folder at your workspace root:

• Chat history in JSON format.
• Codebase analysis database (SQLite).
• Project rules and configuration.
• Logs and session data.

This folder is automatically added to .gitignore to prevent committing sensitive data.

Troubleshooting

Connection Issues with Local Models

• Verify Ollama or LM Studio is running.
• Check the port configuration (default: 11434 for Ollama, 1234 for LM Studio).
• Confirm the local.baseUrl setting matches your server.

Agent Not Responding

• Use the Stop button in the chat interface.
• Clear chat history via Settings > Privacy.
• Check the CodeBuddy output channel (View > Output > CodeBuddy).

API Key Errors

• Verify your API key is correctly entered in settings.
• Ensure the selected model matches your API key provider.
• Check for network connectivity for cloud providers.

Contributing

We are building the future of software engineering. Contributions are welcome. Please see CONTRIBUTING.md for guidelines.

License

MIT License - see LICENSE for details.

---

Repository: github.com/olasunkanmi-SE/codebuddy