forloopcodes/contextplus

What it solves

Context+ provides high-accuracy semantic intelligence for large-scale engineering projects. It solves the problem of navigating and understanding massive codebases where traditional text search is insufficient, allowing developers to map features to code and trace dependencies with high precision.

How it works

It operates as a Model Context Protocol (MCP) server that integrates with AI coding agents (like Claude Code, Cursor, and Windsurf). It uses a combination of several technologies to index and analyze code:

• Structural Analysis: Uses Tree-sitter AST (Abstract Syntax Tree) to parse 43 different languages, allowing it to identify function signatures, class methods, and symbol ranges.
• Semantic Search: Employs embeddings (via Ollama or OpenAI-compatible providers) to enable searching by meaning rather than exact text.
• Graph-Based Memory: Implements a property graph with decay scoring and Obsidian-style "wikilink" hubs to map features to specific files.
• Clustering: Uses spectral clustering to group semantically related files into labeled clusters for easier browsing.
• Safety Mechanisms: Includes a "shadow restore point" system that captures file states before AI-driven changes, allowing for undos without affecting git history.

Who it’s for

Software engineers working on large, complex codebases who use AI coding agents and need a more reliable way to retrieve relevant code context and analyze the "blast radius" of changes.

Highlights

• Deep Code Intelligence: Combines AST parsing with semantic embeddings for identifier-level retrieval.
• MCP Integration: Ready-to-use with major AI IDEs and agents.
• ** uma
• Graph Traversal: Memory nodes can be interlinked with typed edges (e.g., depends_on, implements) and searched via graph traversal.
• Shadow Restore Points: Provides a safety net for AI code modifications with a dedicated undo system.
• Multi-Language Support: Supports a wide range of languages including TypeScript, Python, Rust, and Go.