Skip to content
| Marketplace
Sign in
Visual Studio Code>Programming Languages>Coders in FlowNew to Visual Studio Code? Get it now.
Coders in Flow

Coders in Flow

Coders In Flow

|
9 installs
| (1) | Free
Coders in Flow AI Code Editor
Installation
Launch VS Code Quick Open (Ctrl+P), paste the following command, and press enter.
Copied to clipboard
More Info

Coders in Flow

A powerful VS Code extension for AI-powered coding assistance, featuring a sophisticated context management system inspired by Roo-Code, with extensive multi-provider AI support, semantic code search, and Git integration.

🏗️ Built for Developers

Written from the ground up to be easy to debug and extend:

  • Comprehensive Logging System - Multiple logging channels with ID-based filtering for targeted debugging
  • Modular Architecture - Clean separation of concerns makes adding new features quick and straightforward
  • Provider Plugin System - Adding new AI providers requires implementing just one interface
  • Tool Plugin System - New tools can be added by dropping in a single file
  • Debug Commands - Built-in commands for testing connections, validating configurations, and troubleshooting issues
  • Development Mode - Special logging and debugging features activated during development

🚀 Key Features

🤖 Multi-Provider AI Support

Support for 12+ AI providers with seamless switching:

  • Anthropic Claude (Claude 3 Opus, Sonnet, Haiku)
  • OpenAI (GPT-4, GPT-3.5)
  • Google Gemini (Pro, Flash - with caching support)
  • Claude Code CLI (Direct integration with claude.ai/code)
  • DeepSeek (Coder and Chat models)
  • Mistral AI
  • Ollama (Local models)
  • LM Studio (Local models)
  • OpenRouter (Multi-provider gateway)
  • AWS Bedrock
  • Google Vertex AI
  • VS Code Language Models (GitHub Copilot integration)

🔍 Semantic Code Search

Advanced vector-based code search powered by embeddings:

  • Qdrant Vector Database for fast similarity search
  • Smart Code Chunking with configurable overlap
  • Multiple Embedding Providers (OpenAI is best for cost around $0.2 for 3000 files)
  • Project-Aware Indexing with intelligent file filtering
  • Incremental Indexing for changed files only
  • Instant Project Understanding - AI immediately knows project details via auto-generated summaries
  • No More Context Confusion - AI never mistakes what project it's working on
  • Reduced Search Time - AI spends less time learning about your codebase
  • Commands: semanticSearch, indexWorkspace, indexFile, generateProjectSummary

🗃️ ShadowGit - Advanced Version Control

Built-in Git-like version control system:

  • Automatic File Snapshots with change tracking
  • Checkpoint System for saving and restoring states
  • Visual Diff Viewer with side-by-side comparison
  • Great Git Interface - Intuitive UI for all Git operations
  • Git Integration for syncing with actual Git repositories
  • Staging Support with file-level operations
  • Checkpoint Management UI via dedicated webview panel

💬 Intelligent Chat Interface

React-based chat UI with advanced features:

  • Task-Based Conversations - Organize work into separate tasks
  • Smart Context Management - Intelligent conversation history handling
  • Mode-Based Assistance - Different modes for different tasks
  • Real-time Token Counting with usage tracking
  • Code Preview with syntax highlighting
  • Inline Diff Viewing for code changes
  • Markdown Rendering with full GitHub Flavored Markdown support

⚙️ Admin Interface

Easy-to-use configuration panel for complete customization:

  • Visual Settings Editor - No more JSON editing, use intuitive UI
  • Prompt Customization - Edit system prompts for each mode visually
  • Provider Management - Add/remove/configure AI providers with clicks
  • Model Configuration - Enable/disable models, set preferences
  • Profile Management - Create and switch between multiple profiles
  • Live Preview - See changes in real-time before applying
  • Import/Export - Share configurations with team members

🛠️ Comprehensive Tool System

AI has access to powerful tools for automation:

  • Multiple Tool Calls - Execute multiple tools in a single request to reduce API costs
  • Self-Healing Tool Calls - Automatic correction of malformed requests
  • Parameter Auto-Correction - Fixes spacing, line numbers, and formatting issues
  • Smart Error Recovery - No more failed tool calls from minor AI mistakes
  • File Operations: Read, write, search, list files
  • Code Manipulation: Insert at location, search & replace
  • Command Execution: Run terminal commands
  • Semantic Search: Query codebase by meaning
  • Diagram Creation: Generate Mermaid diagrams
  • Memory Storage: Persistent information storage

🎯 Context Management System

Sophisticated context handling inspired by Roo-Code:

  • Token-Aware Truncation - Smart message prioritization
  • Visual Context Editor - See and edit what's included in context
  • Context Exclusion - Remove specific items to reduce costs
  • System Prompt Optimization - Mode-specific prompts
  • Custom Mode Prompts - Define your own prompts per mode
  • Environment Awareness - Project context inclusion
  • Multi-Strategy Support - Simple, smart, or summarize modes

📊 Additional Features

  • Terminal Integration - Capture and execute commands
  • Mermaid Diagram Support - Create and render diagrams
  • Project Type Detection - Automatic framework detection
  • MCP Hub Support - Model Context Protocol integration
  • Easily insert files with @ tool based on index of codebase, easy search

💰 Cost Optimization & Intelligence

  • Live Cost Calculation - See API costs calculated in real-time as you chat
  • GitHub Copilot Integration - Unlimited GPT-4.1 access for just $10/month
  • DeepSeek Excellence - Use one of the smartest models available for pennies per day
  • Dynamic Model Updates - Automatically fetch latest models when providers add/remove them
  • Automatic Pricing Updates - Real-time pricing information fetched remotely
  • Usage Tracking - Monitor token usage and costs across all providers
  • Smart Batching - Multiple tool calls per request to minimize API costs
  • Context Optimization - Visual editor to exclude unnecessary context

🏛️ Smart Architecture

  • Profile-Based Design - Everything organized under user profiles
  • Hierarchical Structure - Profiles → API Providers → Models
  • Automatic Model Discovery - Models dynamically fetched from providers
  • Remote Configuration - Pricing and capabilities updated automatically
  • Custom Prompts - Define unique prompts for each mode

🧠 AI Intelligence Features

  • AI Memory System - Remembers your preferences, patterns, and project context
  • Project-Aware Rules - Automatically adapts rules based on project type (React, Node.js, Python, etc.)
  • Smart TypeScript Enforcement - Prevents common TS mistakes, drastically reducing lint errors
  • Dynamic Model Switching - AI automatically switches between models based on task complexity to optimize costs
  • Error Prevention - Proactively catches and fixes common coding mistakes before they happen
  • Context-Aware Suggestions - Learns from your codebase to provide better recommendations

Human Cerebrum Mode

  • Coming soon new AI model that thinks like a human to keep context small, while having full information available, lowering costs by large extremes

🚀 Getting Started

Installation

  1. Install from VS Code Marketplace or download the VSIX file
  2. Open VS Code Command Palette (Cmd+Shift+P / Ctrl+Shift+P)
  3. Run Coders in Flow: Open Chat to start

Configuration

  1. Choose your AI provider in settings
  2. Add your API key
  3. Select your preferred model
  4. Configure context window size and strategy

⚙️ Extension Settings

AI Provider Settings

  • codersinflow.ai.provider: Select AI provider (OpenAI, Anthropic, etc.)
  • codersinflow.ai.apiKey: Your API key for the selected provider
  • codersinflow.ai.model: The model to use (e.g., gpt-4, claude-3-opus)

Context Management

  • codersinflow.context.maxTokens: Maximum tokens for context window
  • codersinflow.context.strategy: Truncation strategy (simple, smart, summarize)

Semantic Search

  • codersinflow.search.embeddingProvider: Embedding provider (OpenAI, Xenova)
  • codersinflow.search.chunkSize: Code chunk size for indexing
  • codersinflow.search.chunkOverlap: Overlap between chunks

📚 Usage Examples

Basic Chat

  1. Open the chat panel
  2. Type your question or request
  3. The AI will respond with code suggestions, explanations, or solutions

Semantic Search

1. Index your workspace: Cmd+Shift+P → "Index Workspace"
2. Search semantically: Cmd+Shift+P → "Semantic Search"
3. Enter natural language query like "authentication logic"

ShadowGit Checkpoints

1. Open ShadowGit: Cmd+Shift+P → "Open ShadowGit"
2. Make code changes
3. Create checkpoint with description
4. Restore to any previous checkpoint if needed

Context Management

1. View Context: Cmd+Shift+P → "View Context"
2. See all messages, files, and data included
3. Click items to exclude from context
4. Monitor token count in real-time
5. Save on API costs by removing unnecessary items

🧩 Task Management & Git Integration

How Tasks Work

  • Each task gets its own conversation history
  • Tasks can be paused, resumed, and completed
  • Context is preserved across task switches

Git Branch Behavior

When creating a new task:

  • Creates and checks out a new Git branch
  • Branch name derived from task description

When switching tasks in UI:

  • UI state updates to show active task
  • Git branch remains unchanged
  • Changes go to current branch, not UI-selected task

Important: Be aware that task switching in the UI doesn't switch Git branches. Your changes always go to the current Git branch.

🔧 Development

Building from Source

# Install dependencies
npm install

# Development build with watch
npm run watch

# Production build
npm run package

# Build VSIX package
npm run build

Testing

# Run all tests
npm test

# Run specific test
jest path/to/test.test.ts

# Type checking
npm run check-types

# Linting
npm run lint

Debugging Features

The extension includes extensive debugging capabilities:

Console Logging System

Enable specific log categories by adding IDs to ENABLED_LOG_IDS in src/utils/consoleLog.ts:

const ENABLED_LOG_IDS = [
  'tool-execution',    // Track tool calls
  'message-store',     // Message handling
  'context-manager',   // Context window management
  'provider-api',      // AI provider communication
  // Add more IDs as needed
];

Debug Commands

  • testConnection - Test AI provider connectivity
  • validateApiKey - Verify API key configuration
  • debugContext - View current context window
  • testClaudeCodeTerminal - Test Claude Code integration
  • checkWorkspaceIndexing - Verify semantic search status

Adding New Features

New AI Provider:

// Simply implement the BaseProvider interface
export class MyProvider extends BaseProvider {
  async complete(messages, options) {
    // Your implementation
  }
}

New Tool:

// Add to src/tools/ directory
export const myTool = {
  name: 'myTool',
  description: 'What it does',
  inputSchema: { /* JSON schema */ },
  execute: async (input) => {
    // Tool logic
  }
};

📦 Requirements

  • VS Code 1.74.0 or higher
  • Node.js 20.18.3 or higher (for development)
  • API key for your chosen AI provider

🐛 Known Issues

  • Git branch doesn't auto-switch with task changes
  • Some providers may have connection issues in certain regions
  • Large workspaces may take time to index for semantic search

📄 License

MIT

🤝 Contributing

Contributions are welcome! Please read our contributing guidelines and submit PRs to our GitHub repository.

📮 Support

  • Report issues on GitHub
  • Join our community discussions
  • Check the wiki for detailed documentation

Possible Issues

  • insert content and search_and_replace seem to overlap

APIs

  • https://github.com/googleapis/js-genai

Enjoy coding with AI assistance! 🎉

  • Contact us
  • Jobs
  • Privacy
  • Manage cookies
  • Terms of use
  • Trademarks
© 2025 Microsoft