SkyDeckAI Code

by skydeckai

Not rated
GitHub

About

A comprehensive toolkit for AI-driven development, offering file system operations, code analysis, execution, web searching, and system information retrieval.

Details

Author
skydeckai
Categories
Developer Tools, Other

Setup

Install SkyDeckAI Code in your MCP client (Claude Desktop, Cursor, Windsurf, and others).

Repository: https://github.com/skydeckai/skydeckai-code

Follow the installation instructions in the repository README, then restart your MCP client.

An MCP server that provides a comprehensive set of tools for AI-driven development workflows. Features include file system operations, code analysis using tree-sitter for multiple programming languages, code execution, web content fetching with HTML-to-markdown conversion, multi-engine web search, code content searching, and system information retrieval. Designed to enhance AI's capability to assist in software development tasks by providing direct access to both local and remote resources.

This mcp server was formerly known asmcp-server-aidd. It was renamed toskydeckai-codeto credit the team atSkyDeck.aiwith creating this application along withEast Agile. But more importantly we realized that the term AI Driven Development (AIDD) was just not catching on. People did not understand at a glance what it was about. And nor did LLMs. "Code" was far more intuitive. And linguistically intuitive is important in the world of agentic AI.

Add to yourclaude_desktop_config.json:

{ "mcpServers": { "skydeckai-code": { "command": "uvx", "args": ["skydeckai-code"] } } }

If you're using MseeP AI Helper app, you can search for "SkyDeckAI Code" and install it.

- File system operations (read, write, edit, move, copy, delete)
- Directory management and traversal
- Multi-language code analysis using tree-sitter
- Code content searching with regex pattern matching
- Multi-language code execution with safety measures
- Web content fetching from APIs and websites with HTML-to-markdown conversion
- Multi-engine web search with reliable fallback mechanisms
- Batch operations for parallel and serial tool execution
- Security controls with configurable workspace boundaries
- Screenshot and screen context tools
- Image handling tools

Pattern-based file editing with preview support:

{ "path": "src/main.py", "edits": [ { "oldText": "def old_function():", "newText": "def new_function():" } ], "dryRun": false, "options": { "partialMatch": true } }

Returns: Diff of changes or preview in dry run mode.

Thesearch_filestool searches for files by name pattern, while thesearch_codetool searches within file contents using regex. Usesearch_fileswhen looking for files with specific names or extensions, andsearch_codewhen searching for specific text patterns inside files.

{ "path": "src", "include_hidden": false }

Returns: JSON tree structure of directory contents.

- Classes and their methods
- Functions and parameters
- Module structure
- Code organization statistics
- Inheritance relationships

- Python (.py)
- JavaScript (.js/.jsx, .mjs, .cjs)
- TypeScript (.ts/.tsx)
- Java (.java)
- C++ (.cpp, .hpp, .cc)
- Ruby (.rb, .rake)
- Go (.go)
- Rust (.rs)
- PHP (.php)
- C# (.cs)
- Kotlin (.kt, .kts)

Fast content search tool using regular expressions:

{ "patterns": ["function\\s+\\w+", "class\\s+\\w+"], "include": ".js", "exclude": "node_modules/", "max_results": 50, "case_sensitive": false, "path": "src" }

Returns:Matching lines grouped by file with line numbers, sorted by file modification time with newest files first.

This tool uses ripgrep when available for optimal performance, with a Python fallback implementation. It's ideal for finding specific code patterns like function declarations, imports, variable usages, or error handling.

{ "working_directory": "/path/to/project", "system": { "os", "os_version", "architecture", "python_version" }, "wifi_network": "MyWiFi", "cpu": { "physical_cores", "logical_cores", "total_cpu_usage" }, "memory": { "total", "available", "used_percentage" }, "disk": { "total", "free", "used_percentage" }, "mac_details": { // Only present on macOS "model": "Mac mini", "chip": "Apple M2", "serial_number": "XXX" } }

Provides essential system information in a clean, readable format.

Returns a list of currently active applications on the user's system.

{ "success": true, "platform": "macos", "app_count": 12, "apps": [ { "name": "Firefox", "has_windows": true, "window_count": 3, "visible_windows": [ { "name": "GitHub - Mozilla Firefox", "width": 1200, "height": 800 } ] }, { "name": "VSCode", "has_windows": true } ] }

This tool provides valuable context about applications currently running on the user's system, which can help with providing more relevant assistance.

Returns detailed information about all available windows currently displayed on the user's screen.

{ "success": true, "platform": "macos", "count": 8, "windows": [ { "id": 42, "title": "Document.txt - Notepad", "app": "Notepad", "visible": true }, { "title": "Terminal", "app": "Terminal", "visible": true, "active": true } ] }

This tool helps understand what's visible on the user's screen and can be used for context-aware assistance.

Captures a screenshot of the user's screen or a specific window.

{ "output_path": "screenshots/capture.png", "capture_mode": { "type": "named_window", "window_name": "Visual Studio Code" } }
{ "success": true, "path": "/path/to/screenshots/capture.png" }

This tool captures screenshots for visualization, debugging, or context-aware assistance.

Reads an image file from the file system and returns its contents as a base64-encoded string.

{ "path": "images/logo.png" }

Returns:Base64-encoded image data that can be displayed or processed.

This tool supports common image formats like PNG, JPEG, GIF, and WebP, and automatically resizes images for optimal viewing.

Fetches content from a URL and optionally saves it to a file.

{ "url": "https://api.github.com/users/octocat", "headers": { "Accept": "application/json" }, "timeout": 15, "save_to_file": "downloads/octocat.json", "convert_html_to_markdown": true }

Returns:Response content as text with HTTP status code and size information. For binary content, returns metadata and saves to file if requested. When convert_html_to_markdown is enabled, HTML content is automatically converted to markdown format for better readability.

This tool can be used to access web APIs, fetch documentation, or download content from the web while respecting size limits (10MB max) and security constraints.

Performs a robust web search using multiple search engines and returns concise, relevant results.

{ "query": "latest python release features", "num_results": 8, "convert_html_to_markdown": true, "search_engine": "bing" }

Returns:A list of search results formatted in markdown, including titles, URLs, and snippets for each result. Results are deduplicated and organized hierarchically for easy reading.

This tool uses a multi-engine approach that tries different search engines with various parsing strategies to ensure reliable results. You can specify a preferred engine, but some engines may block automated access, in which case the tool will fall back to alternative engines when "auto" is selected.

Execute multiple tool invocations in a single request with parallel execution when possible.

{ "description": "Setup new project", "sequential": true, "invocations": [ { "tool": "create_directory", "arguments": { "path": "src" } }, { "tool": "write_file", "arguments": { "path": "README.md", "content": "# New Project\n\nThis is a new project." } }, { "tool": "execute_shell_script", "arguments": { "script": "git init" } } ] }

Returns:Combined results from all tool invocations, grouped by tool with success/error status for each. Results are presented in the original invocation order with clear section headers.

This tool provides efficient execution of multiple operations in a single request. Whensequentialis false (default), tools are executed in parallel for better performance. Whensequentialis true, tools are executed in order, and if any tool fails, execution stops.

IMPORTANT: All tools in the batch execute in the same working directory context. If a tool creates a directory and a subsequent tool needs to work inside that directory, you must either:
- Use paths relative to the current working directory (e.g., "project/src" rather than just "src"), or
- Include an explicit tool invocation to change directories usingupdate_allowed_directory

A tool for complex reasoning and brainstorming without making changes to the repository.

{ "thought": "Let me analyze the performance issue in the codebase:\n\n## Root Cause Analysis\n\n1. The database query is inefficient because:\n - It doesn't use proper indexing\n - It fetches more columns than needed\n - The JOIN operation is unnecessarily complex\n\n## Potential Solutions\n\n1. Add database indexes:\n - Create an index on the user_id column\n - Create a composite index on (created_at, status)\n\n2. Optimize the query:\n - Select only necessary columns\n - Rewrite the JOIN using a subquery\n - Add LIMIT clause for pagination\n\n3. Add caching layer:\n - Cache frequent queries using Redis\n - Implement cache invalidation strategy\n\nAfter weighing the options, solution #2 seems to be the simplest to implement with the highest impact." }

Returns:Your thoughts formatted as markdown, with a note indicating this was a thinking exercise.

This tool is useful for thinking through complex problems, brainstorming solutions, or laying out implementation plans without making any actual changes. It's a great way to document your reasoning process, evaluate different approaches, or plan out a multi-step strategy before taking action.

Executes code in various programming languages with safety measures and restrictions.

{ "language": "python", "code": "print('Hello, World!')", "timeout": 5 }

- Python (python3)
- JavaScript (Node.js)
- Ruby
- PHP
- Go
- Rust

- Respective language runtimes must be installed
- Commands must be available in system PATH
- Proper permissions for temporary file creation

⚠️Security Warning:This tool executes arbitrary code on your system. Always:
- Review code thoroughly before execution
- Understand the code's purpose and expected outcome
- Never execute untrusted code
- Be aware of potential system impacts
- Monitor execution output

Executes shell scripts (bash/sh) with safety measures and restrictions.

{ "script": "echo \"Current directory:\" && pwd", "timeout": 300 }

- Uses /bin/sh for maximum compatibility across systems
- Executes within the allowed directory
- Separate stdout and stderr output
- Proper error handling and timeout controls

⚠️Security Warning:*This tool executes arbitrary shell commands on your system. Always:
- Review the script thoroughly before execution
- Understand the script's purpose and expected outcome
- Never execute untrusted scripts
- Be aware of potential system impacts
- Monitor execution output

The todo tools provide sequential task management capabilities for workspace-first development workflows. Tasks are executed in order without priority systems, ensuring structured progress through development phases.

Read the current todo list for the workspace.

{ "todos": [ { "id": "abc123", "content": "Implement user authentication", "status": "in_progress", "metadata": { "custom_key": "custom_value" }, "created_at": "2023-10-01T10:00:00Z", "updated_at": "2023-10-01T11:30:00Z" } ], "count": 1, "workspace": "/path/to/workspace" }

Replace the entire todo list for sequential execution workflow. Tasks are executed in array order, building upon previous work.

{ "todos": [ { "id": "task1", "content": "Set up database schema", "status": "pending" }, { "id": "task2", "content": "Create API endpoints", "status": "pending", "metadata": { "custom_key": "custom_value" } } ] }

- Each todo must have unique ID
- Only one task can be "in_progress" at a time (sequential execution)
- Tasks execute in array order - no priority system
- Required fields: id, content, status
- Status values: "pending", "in_progress", "completed"
- Workspace-first: Todo management is mandatory for all workspace operations

Update a specific todo item by ID for sequential workflow progression.

{ "todo_id": "task1", "updates": { "status": "in_progress", "metadata": { "new_key": "new_value" } } }
{ "success": true, "updated_todo": { "id": "task1", "content": "Set up database schema", "status": "in_progress", "updated_at": "2023-10-01T12:00:00Z", "metadata": { "new_key": "new_value" } }, "counts": { "pending": 1, "in_progress": 1, "completed": 0, "total": 2 }, "workspace": "/path/to/workspace" }

The todo system maintains separate sequential task lists for each workspace, enforcing mandatory usage for all workspace operations. Tasks execute in order, building upon previous work without priority-based scheduling.

Configuration file:~/.skydeckai_code/config.json

{ "allowed_directory": "/path/to/workspace" }

- Operations restricted to configured allowed directory
- Path traversal prevention
- File permission preservation
- Safe operation handling

- GitHub tools:

- PR Description Generator
- Code Review
- Actions Manager

Currently in active development. Features and API may change.

This is a web browser that enables your coding agent, such as Claude Code, to visit websites on your behalf and assist you in identifying bugs or creating UI test cases.

No reviews yet — be the first

Sign in to leave a review

Use Google, GitHub, or an email account so ratings stay tied to real people.

Email sign in

No reviews posted yet.