This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
The SPARC Automated Development System (claude-sparc.sh) is a comprehensive, agentic workflow for automated software development using the SPARC methodology (Specification, Pseudocode, Architecture, Refinement, Completion). This system leverages Claude Code's built-in tools for parallel task orchestration, comprehensive research, and Test-Driven Development.
We propose an AI agent architecture in PyTorch that integrates state-of-the-art components to meet the following goals: (1) advanced reasoning with transformer models, (2) ingestion of large documents or histories via long context windows, (3) persistent memory without traditional vector-database RAG, (4) tool use for actions (API calls, code execution, etc.) similar to Anthropic’s MCP standard, and (5) declarative, goal-driven behavior with autonomous planning. The system will be compatible with both CPU and GPU environments. Below, we detail recommended models, libraries, and design choices for each aspect, followed by an overall architecture and example implementation steps.
Model Selection: Use modern transformer-based LLMs known for strong reasoning and multitasking. For example, Meta’s LLaMA 2 (open-source, 7B–70B parameters) or **Mist
The OpenAI Codex Machine Learning Setup is a comprehensive environment designed for building advanced AI-powered applications with a focus on agentic capabilities. This project provides a robust foundation for creating AI systems that can reason about complex tasks, interact with external tools, and execute actions on behalf of users.
Built around a core set of modern AI libraries and tools, this setup enables the development of sophisticated machine learning pipelines, particularly those leveraging Large Language Models (LLMs) for reasoning and decision-making. The project structure integrates seamlessly with FastMCP for standardized API interfaces and provides connectivity with various external services through tool integrations.
Environment Configuration
.env file should be included in .gitignore to prevent committing sensitive information like API keys. This is mentioned in the README.md, but it must be enforced.Database Files
agent_registry.db file is skipped in commits, but should be checked to ensure it doesn't contain sensitive information or credentials.Key Management
src/app/api/secure-binding/ca/route.ts stores CA keys in memory. Not secure for production. Use a secure key management service.
A comprehensive guide to building an AI-native IDE inspired by Windsurf and Cursor using VSCode and Roo Code
The rise of AI-native IDEs like Windsurf (formerly Codeium) and Cursor has redefined developer productivity. These tools integrate AI agents with deep codebase understanding, collaborative workflows, and streamlined coding experiences. While Windsurf and Cursor are standalone applications, developers can create similar solutions by leveraging Roo Code-an open-source VSCode extension-and building a custom VSCode distribution.
This guide outlines the steps to create rUv Code, a tailored VSCode distribution centered around Roo Code’s AI capabilities, with features comparable to commercial AI IDEs.
| { | |
| "customModes": [ | |
| { | |
| "slug": "fire-crawler", | |
| "name": "🔥 Fire Crawler", | |
| "roleDefinition": "You are a specialized web crawling and data extraction assistant that leverages Firecrawl to gather, analyze, and structure web content. You extract meaningful information from websites, perform targeted searches, and create structured datasets from unstructured web content.", | |
| "customInstructions": "You use Firecrawl's advanced web crawling and data extraction capabilities to gather and process web content efficiently. You:\n\n• Crawl websites recursively to map content structures\n• Extract structured data using natural language prompts or JSON schemas\n• Scrape specific content from web pages with precision\n• Search the web and retrieve full page content\n• Map website structures and generate site maps\n• Process and transform unstructured web data into usable formats\n\n## Web Crawling Strategies\n\n1. **Site Mapping**: Use FIRECRAWL_MAP_URLS to discover and map website structures\n2. ** |
| { | |
| "slug": "deep-research", | |
| "name": "🔍 Deep Research Mode", | |
| "roleDefinition": "You are a specialized research assistant that leverages Perplexity AI to conduct deep, comprehensive research on any topic, creating structured documentation and reports through a recursive self-learning approach.", | |
| "customInstructions": "You use Perplexity AI's advanced search capabilities to retrieve detailed, accurate information and organize it into a comprehensive research documentation system writing to a research sub folder and final report sub folder with ToC and multiple md files. You:\n\n• Craft precise queries to extract domain-specific information\n• Provide structured, actionable research with proper citations\n• Validate information across multiple sources\n• Create a hierarchical documentation structure\n• Implement recursive self-learning to refine and expand research\n\n## Research Documentation Structure\n\nFor each research project, create the following folder structure:\n\n```\nresearch/\n |
| Introduction | |
| Building a local serverless runtime for agent command/control systems involves using WebAssembly (WASM) modules as secure, ephemeral plugins executed via a command-line interface (CLI). In this architecture, each agent command is implemented as an isolated WASM module (e.g. compiled from Rust or AssemblyScript) that the agent can invoke on-demand. This approach leverages WebAssembly’s strengths – near-native performance, cross-platform portability, and strong sandboxing – to ensure commands run efficiently and safely on any host  . By treating each CLI action as a “function-as-a-service” invocation, we achieve a local serverless model: commands execute on demand with no persistent runtime beyond their execution. The plan outlined below covers the full implementation details, from toolchain and CLI design to security, performance, and integration with the Model Context Protocol (MCP) for orchestrating multiple agents. | |
| High-Level Design: A central Controller (which could be an MCP client or or |
Powered by composio this MCP.json provides detailed information on Model Context Protocol (MCP) integration capabilities and enables seamless agent workflows by connecting to more than 80 servers.
It covers development, AI, data management, productivity, cloud storage, e-commerce, finance, communication, and design. Each server offers specialized tools, allowing agents to securely access, automate, and manage external services through a unified and modular system. This approach supports building dynamic, scalable, and intelligent workflows with minimal setup and maximum flexibility.