Universal AI context generator that significantly reduces token consumption for AI coding assistants by providing pre-digested project knowledge. It targets developers using tools like Claude Code, Cursor, and Copilot, offering substantial efficiency gains and cost savings by eliminating repetitive context-gathering by the AI.

How It Works

CodeSight leverages Abstract Syntax Tree (AST) parsing for TypeScript projects and robust regex detection for others to precisely analyze code structure. It identifies key elements like API routes, database schemas, UI components, and dependencies, then synthesizes this information into structured markdown files. This pre-processed context allows AI assistants to access project-specific knowledge efficiently, drastically cutting down the tokens required for exploration and understanding. The --wiki mode further enhances this by generating a persistent, narrative knowledge base from the codebase.

Quick Start & Requirements

• Primary install/run command: npx codesight
• Prerequisites: Node.js >= 18. Zero external dependencies; it utilizes the project's local TypeScript installation for AST parsing.
• Links: npm package, GitHub repository

Highlighted Details

• Supports 30+ frameworks, 13 ORM parsers, and 13 MCP tools across TypeScript, JavaScript, Python, Go, Ruby, Elixir, Java, Kotlin, Rust, PHP, Dart, Swift, and C#.
• Generates a persistent codebase knowledge base (.codesight/wiki/) that survives across AI sessions, significantly reducing per-session token load.
• The --mode knowledge feature maps external markdown sources (like Obsidian vaults) into an AI context primer (KNOWLEDGE.md).
• Provides detailed "blast radius" analysis to identify transitively affected files, routes, and models before code changes.
• Benchmarks claim up to 131x token reduction with the targeted wiki queries compared to manual exploration.
• Automatically generates AI tool-specific instruction files (e.g., CLAUDE.md, .cursorrules) for immediate context integration.
• Offers an MCP server for on-demand, cached context retrieval by AI tools, minimizing redundant scans.

Maintenance & Community

Developed by Kailesk Khumar, solo founder of houseofmvps.com. The project is actively maintained, with links to the author's professional profiles available. No dedicated community channels (e.g., Discord, Slack) are explicitly mentioned.

Licensing & Compatibility

• License: MIT.
• Compatibility: The MIT license permits commercial use and integration within closed-source projects.

Limitations & Caveats

AST precision is dependent on the target project having TypeScript installed in its node_modules. Static analysis may not fully capture routes defined through highly dynamic patterns (e.g., url.match(/pattern/) within handlers). Regex fallback for non-TypeScript projects may offer less precision than AST-based analysis.