How skydeckai/skydeckai-code works
Project type and purpose
• This is a Python-based MCP server packaged as a CLI tool (“skydeckai-code-cli”) that exposes ~26 “tools” for AI-driven software development workflows.
• Its purpose is to let an AI agent invoke file-system operations, code analysis (via tree-sitter), code execution, web fetching/searching, screenshot capture, system info, etc., through a common MCP protocol.
File structure
• Top-level:
• Dockerfile, pyproject.toml, smithery.yaml (packaging for MCP), README, LICENSE
• src/aidd/cli.py (entrypoint CLI)
• src/aidd/server.py (MCP server that registers tools)
• src/aidd/tools/… (each tool as a Python module)
• Pattern: a single process hosting a tool-registry and dispatching invocations to individual tool modules.
• Configuration: looks for ~/.skydeckai_code/config.json for workspace restrictions.
README for additional insights
• Lists the 26 tools in categories (File System, Code Tools, Web Tools, Screen Tools, System, Utility).
• Describes external dependencies: tree-sitter for parsing, ripgrep fallback, HTML-to-Markdown conversion, multiple search engines, screenshot libs, image handling.
• Shows CLI usage and security constraints (allowed directory sandboxing).
Elements
• Blue for internal components (CLI, Server, Tools)
• Green for external integrations (FS, Network, Parser, Screenshot)
• Orange for configuration and packaging artifacts (config file, Docker)
Final advice
• Sandbox boundary (allowed_directory).
• Technology callouts (tree-sitter, ripgrep, PIL, requests).