Codistry

The Future of Development

Codistry is a privacy-first AI coding assistant for Visual Studio Code. Connect to your own AI providers, local models, or corporate endpoints - your code never touches our servers.

🔒 Privacy First

Your code stays yours. Codistry runs entirely within VS Code and connects directly to AI providers you control:

• Bring Your Own API Keys - Use your Anthropic or OpenAI accounts
• Local Models - Run completely offline with Ollama, LM Studio, or other local LLMs
• Corporate Endpoints - Connect to your company's AI infrastructure
• Zero Data Collection - We never see your code, prompts, or conversations

Perfect for enterprise environments where data privacy is critical.

✨ Key Features

🤖 Connect to Any AI Provider

• Anthropic Claude - Use your API key for Claude Sonnet 4.5 and other models
• OpenAI - Connect with your API key for GPT-5, GPT-4o, o1, o3, and more
• Custom Endpoints - Point to any OpenAI-compatible API (Ollama, LM Studio, corporate deployments)
• Fully Configurable - Multiple endpoints, custom model definitions, fine-tuned control

💡 Intelligent Code Understanding

• Code Navigation - Jump to definitions, find references, browse symbols
• Real-time Diagnostics - AI sees linting errors, warnings, and type issues
• Smart Code Editing - Context-aware file editing
• Symbol Renaming - Safely refactor across your entire codebase

🔧 Powerful Development Tools

• File Operations - Read, search, and edit files with AI assistance
• Grep Search - Find patterns across your entire workspace
• Terminal Commands - Execute commands and run tasks directly from chat
• VS Code Integration - Access any VS Code command through the agent

📊 Visual Insights

• Diagram Generation - Create Mermaid diagrams directly in chat
• Step-by-Step Tracking - Watch what the agent is doing in real-time
• Multi-workspace Support - Work across multiple workspace folders seamlessly

🧠 Advanced AI Features

• Reasoning Models - Support for OpenAI's o1 o3, and GPT-5 with configurable effort levels
• Project Memory - AI builds a knowledge graph about your codebase
• Context Rules - Define custom guidelines for AI behavior in your project
• Conversation Management - Automatic summarization or sliding window context

🌿 Chat Worktrees

• Isolated checkouts per task - Start a chat in its own git worktree so parallel work never touches your checkout (/worktree <task>, or /worktree start <task> when the task itself begins with a lifecycle word such as ship)
• Rebase, ship, clean up from chat - /worktree rebase, /worktree ship (commit and push, then offer to create a PR — via gh or a handoff to your Git host; the editable title defaults to the first pushed commit's subject or the branch, the description to one bullet per pushed commit, and a pasted PR URL is recorded for the pushed repository only), /worktree cleanup
• Same lifecycle in the terminal - /worktree … and /worktrees in the codistry CLI, plus codistry worktree list|start|rebase|ship|cleanup … [--json] [--yes] for scripts

🔌 Extensibility

• MCP Integration - Connect to external tools via Model Context Protocol
• Auto-approval - Configure which tools run automatically vs. need confirmation
• Custom Endpoints - Use your own API URLs for any provider

🚀 Getting Started

1. Install the Extension

   • Search for "Codistry" in the VS Code marketplace
   • Or install from the Extensions panel (Ctrl/Cmd + Shift + X)

2. Configure Your AI Provider

   • Click the Codistry icon in the sidebar
   • Open Settings (gear icon)
   • Choose your setup:
     • Cloud: Add your Anthropic or OpenAI API key
     • Local: Point to your Ollama, LM Studio, or other local model endpoint
     • Corporate: Configure your company's AI service URL

3. Start Chatting

   • Ask the agent to help with your code
   • Request file edits, code reviews, or explanations
   • Let it search your codebase and make changes

💬 Example Use Cases

"Refactor this function to use async/await"

• Agent reads the file, understands context, and suggests or applies changes

"Find all files that import UserService"

• Agent searches your codebase and provides a comprehensive list

"Add error handling to the API routes"

• Agent navigates your code, identifies issues, and implements fixes

"Create a sequence diagram for the authentication flow"

• Agent analyzes your code and generates a Mermaid diagram

"What's causing this TypeScript error?"

• Agent checks diagnostics and explains the issue with suggested fixes

⚙️ Configuration Options

Provider Settings

• Switch between Anthropic, OpenAI, or custom providers
• Configure API endpoints for self-hosted models
• Set reasoning effort for o1/o3 models (low/medium/high)

Behavior Settings

• Max Iterations - Control how many steps the agent can take
• Auto-approve Tools - Decide which tools run without confirmation
• Context Rules - Add project-specific instructions to every request
• Streaming - Enable/disable response streaming per provider

Advanced Features

• Project Memory - Let AI build persistent knowledge about your code
• MCP Servers - Connect additional tools and services
• Custom Models - Define your own OpenAI-compatible endpoints

🔐 Security & Privacy

Built for Enterprise and Sensitive Codebases

• Zero Data Collection - Codistry never receives your code, prompts, or API keys
• Direct Connections - Extension connects directly from your machine to your chosen provider
• Secure Storage - API keys stored in VS Code's encrypted secret storage
• On-Device Options - Use local models for completely offline operation
• Tool Approval - Control which operations can run automatically (configurable per tool)
• Corporate Ready - Connect to your company's AI infrastructure behind firewalls

Use at work without risking data leaks. Perfect for proprietary codebases, regulated industries, and security-conscious teams.

Made with ❤️ by Adronite

Ready to supercharge your development workflow? Get started now!