Skip to content
| Marketplace
Sign in
Visual Studio Code>Programming Languages>M.I.K.E. - Machine-Integrated Koding ExtensionNew to Visual Studio Code? Get it now.
M.I.K.E. - Machine-Integrated Koding Extension

M.I.K.E. - Machine-Integrated Koding Extension

Michael Dziejma

| (0) | Free
Production-ready autonomous AI coding assistant for VS Code powered by OpenAI-compatible LLM endpoints, local engines (Ollama, vLLM), and Poolside
Installation
Launch VS Code Quick Open (Ctrl+P), paste the following command, and press enter.
Copied to clipboard
More Info

M.I.K.E. Logo

M.I.K.E. — Machine-Integrated Koding Extension

VS Marketplace Version VS Marketplace Downloads Open VSX Version Open VSX Downloads VS Code License Zero Dependencies LLM Compatibility

M.I.K.E. (Machine-Integrated Koding Extension) is a high-performance, autonomous AI software engineering extension for Visual Studio Code. Built natively with the VS Code Extensibility API and modern Web Standards (ES2022 / Node 18+ global fetch), M.I.K.E. works seamlessly with any OpenAI-compatible LLM endpoint — including local open-weight inference servers (Ollama, LM Studio, vLLM, LocalAI), cloud APIs (OpenAI, Groq, LiteLLM), and enterprise private inference clusters.


⚡ 60-Second Quickstart & Initialization

1. Install Extension

Install from the Open VSX Registry, VS Code Marketplace, or via .vsix:

code --install-extension mike-koding-extension-2.3.0.vsix --force

2. Connect to Your LLM Backend

Click the M.I.K.E. sparkle icon in the VS Code Activity Bar (left), open ⚙️ Config, and choose your provider:

Backend Base URL API Key Model Example
Ollama (Local) http://localhost:11434/v1 (Leave blank) qwen2.5-coder:32b, deepseek-coder-v2
LM Studio (Local) http://localhost:1234/v1 (Leave blank) qwen2.5-coder-7b-instruct
vLLM (Local / Server) http://localhost:8000/v1 (Optional token) Qwen/Qwen2.5-Coder-32B-Instruct
OpenAI (Cloud) https://api.openai.com/v1 sk-... gpt-4o, gpt-4o-mini
LiteLLM / Custom Proxy http://localhost:4000/v1 sk-... Any routed model

3. Test & Auto-Discover Models

Click Test or 🔄 Auto-Discover in the config panel. M.I.K.E. will automatically probe /models, discover all active models on your server, verify connectivity, and save confirmed credentials globally.

4. Start Coding!

  • Sidebar Chat: Ask questions, request file changes, attach @editor or @selection context chips.
  • In-Editor Inline Transform: Highlight code and press Cmd+I (macOS) or Ctrl+I (Windows/Linux).
  • Skill Invocation: Type / in the prompt box to autocomplete discovered workspace and global automation skills.

🌟 Core Superpowers & Features

⚡ 1. In-Editor Inline Transform (Cmd + I / Ctrl + I)

Refactor and transform code directly in the active editor without context-switching to the sidebar:

  • Instant HUD with quick recipes: Optimize Performance, Add Error Handling, Add TypeScript Types & JSDoc, Convert to Async/Await, Generate Unit Tests, or Custom Instruction.
  • Streamed in-memory edits directly into the editor buffer with native Cmd+Z undo support.
  • One-click ✓ Keep, ⏪ Revert, or ✨ Fix Diagnostics.

⏪ 2. Atomic Pre-Write Checkpoints & "Reject All" Rollback

  • Safety Net: Before any AI operation writes to disk, M.I.K.E. captures the exact pre-modification binary state.
  • Side-by-Side Diff View: Automatically launches VS Code's native side-by-side diff (mike-diff://...) for visual verification.
  • Session Changes Bar: Tracks the total number of modified and created files with single-click ⏪ Reject All rollback.

💬 3. Multi-Thread Persistent History

  • Multi-Thread Session Drawer: Click 💬 History to browse, switch, or clean past conversations.
  • Auto-Derived Titles: Threads automatically name themselves based on the user's initial prompt.
  • Workspace State Persistence: Sessions survive VS Code restarts and window reloads.

🩺 4. Language Server Diagnostics & Compiler Self-Healing

  • Live Compiler Feedback: Hooks into active VS Code Language Servers (TypeScript, Python Pylance, C# Roslyn, Go, Rust).
  • Autonomous Error Correction: Detects compiler syntax and type errors immediately after writing code and automatically self-heals in the next turn.
  • Editor Quick-Fix Lightbulb: Click the lightbulb on any squiggly error line $\rightarrow$ ✨ Ask M.I.K.E. to fix: <error>.

🔎 5. Fast Workspace Grep & AST Symbol Search

  • grep_search Tool: Fast, zero-dependency VFS text searching with regex, case-insensitivity, and glob filtering.
  • find_symbol Tool: Queries VS Code's native symbol index for classes, functions, and interfaces in milliseconds.

⚡ 6. Extensible Skill System (/<skill-name>)

  • Automatically discovers and parses SKILL.md workflows from .agent/skills/, .skills/, skills/, and global ~/.gemini/config/skills/ or ~/.config/mike/skills/.
  • Type / in the prompt input to filter and auto-complete skills with full YAML metadata.

🛡️ Zero-Shell Security Architecture

Traditional AI coding plugins often execute subprocess shellouts (echo, cat, Set-Content, powershell.exe, bash, git) to mutate code on disk. In corporate, Windows, or air-gapped workstations, this creates severe failure modes:

  • Process spawn blocks by enterprise endpoint security / AppLocker.
  • PowerShell pipeline character encoding corruptions (CRLF, BOM, UTF-16 bugs).
  • Headless terminal buffer hangs and process deadlocks.

The M.I.K.E. Guarantee:

  1. Pure Virtual File System (VFS): All workspace inspections and mutations are executed strictly via vscode.workspace.fs (writeFile, readFile, readDirectory). Zero shell calls are used for file I/O.
  2. Native Host Networking: Built on Node 18+ global fetch to inherit OS root certificates, corporate proxies, and custom trust stores without third-party npm packages.
  3. Encrypted Secret Storage: API keys are stored in the operating system's Credential Keychain via vscode.SecretStorage.
  4. Zero Runtime Dependencies: Ultra-lightweight bundle size (< 50 KB).

⚙️ Configuration Reference

Configure under Settings (Ctrl+, or Cmd+, $\rightarrow$ Search M.I.K.E.) or via .vscode/settings.json:

Setting Type Default Description
mike.baseUrl string http://localhost:11434/v1 Base URL for OpenAI-compatible inference server.
mike.apiKey string "" API Key / Bearer token (stored securely in OS Keychain).
mike.model string qwen2.5-coder Model ID for completions and tool use.
mike.temperature number 0.0 Sampling temperature (0.0 for deterministic, typo-free code).
mike.maxTokens number 8192 Maximum output token limit per turn.
mike.commandMode string "prompt" Shell execution policy ("prompt", "auto", "deny").
mike.customSkillPaths string[] [] Additional directories to scan for SKILL.md files.

🛠 Available Agent Tools

Tool Parameters Description
write_file path: string, content: string Writes UTF-8 content via vscode.workspace.fs. Triggers in-memory diff and checkpoint.
read_file path: string Reads and returns full UTF-8 file content via vscode.workspace.fs.
list_dir path?: string Lists directory entries (files, directories, symlinks) recursively.
grep_search query: string, isRegex?: boolean, caseInsensitive?: boolean, filePattern?: string Searches workspace files for matching patterns and returns line snippets.
find_symbol query: string Queries VS Code's symbol index for classes, methods, and functions.
get_diagnostics path?: string Returns active Language Server errors and warnings across the workspace.
run_command command: string Executes terminal commands according to user security policy.

📚 In-Depth Documentation

  • 📖 User & Operator Guide — Comprehensive guide for daily workflows and shortcuts.
  • 🧭 Initialization & Lifecycle Guide — Deep-dive into M.I.K.E.'s 6-phase initialization lifecycle.
  • 📦 Extension Publishing Guide — Guide for publishing to Open VSX and VS Code Marketplace.
  • 🏛 Architectural Specifications — Detailed internal architecture and component design.
  • 📋 Functional Requirements Spec — Complete requirements specification.

🚀 Building & Contributing

# Clone repository
git clone https://github.com/<your-org>/mike-koding-extension.git
cd mike-koding-extension

# Install dependencies
npm install

# Compile TypeScript
npm run compile

# Run automated test suite
npm test

# Build VSIX package
npm run package

# Run local mock SSE LLM server for testing
npm run mock-server

📄 License

MIT License. Copyright (c) 2026 M.I.K.E. LLC. See LICENSE for details.

  • Contact us
  • Jobs
  • Privacy
  • Manage cookies
  • Terms of use
  • Trademarks
  • Your Privacy Choices
  • Consumer Health Privacy
© 2026 Microsoft