Release 1.9.0: Code Indexing & Smart Token Savings

We’re excited to announce Remembrances MCP 1.9.0, a feature-packed release that brings two major capabilities: a powerful Code Indexing System for semantic code search and a new how_to_use help system that dramatically reduces token consumption.

🔍 Code Indexing System

The headline feature of this release is the Code Indexing System – a complete solution for AI agents to understand, search, and navigate codebases using semantic search.

What Can You Do?

Search Code by Meaning: Ask for “user authentication and password validation” and find relevant login functions, password checkers, and security modules – even if they don’t contain those exact words.

Navigate Large Codebases: Get instant overviews of file structures, find all implementations of an interface, track down references to a function, and understand call hierarchies.

Manipulate Code Intelligently: Retrieve symbol implementations, replace function bodies, and insert new code at specific locations with full context awareness.

14+ Languages Supported

We’ve integrated tree-sitter for accurate AST parsing across a wide range of languages:

• Go, Rust, C/C++ – Systems programming
• TypeScript, JavaScript – Web development
• Python, Ruby, PHP – Scripting languages
• Java, C#, Kotlin, Scala – Enterprise languages
• Swift – Mobile development
• And more!

How It Works

1. Index your project:

   code_index_project({
     "project_path": "/path/to/project",
     "project_name": "My App"
   })
   

2. Search semantically:

   code_semantic_search({
     "project_id": "my-app",
     "query": "database connection pooling"
   })
   

3. Find and navigate symbols:

   code_find_symbol({
     "project_id": "my-app",
     "name_path_pattern": "DatabasePool/getConnection"
   })
   

The indexer extracts all meaningful symbols – classes, functions, methods, interfaces – and creates vector embeddings for semantic similarity search. Changes are tracked and re-indexed automatically.

💡 Intelligent Help System (how_to_use)

With 37+ tools available, loading full documentation at the start of every conversation consumed ~15,000+ tokens before any actual work began. That’s expensive and inefficient.

The Solution: On-Demand Documentation

The new how_to_use tool provides documentation exactly when you need it:

How It Works

Each tool now has a minimal 1-2 line description. When your AI agent needs more information:

how_to_use("code_semantic_search")

This loads only the documentation for that specific tool – full parameter descriptions, examples, and related tools.

You can also get category overviews:

how_to_use("code")      # All code indexing tools
how_to_use("memory")    # All memory tools
how_to_use("kb")        # All knowledge base tools

Why This Matters

• Lower costs: Fewer tokens per conversation means lower API bills
• Faster responses: Less context to process means quicker initial responses
• Better focus: AI agents see relevant documentation when they need it

Getting Started

Upgrade

Download the latest release from GitHub Releases and replace your existing binary.

Try Code Indexing

1. Start Remembrances MCP
2. Ask your AI to index a project:

   "Index my project at /path/to/project"
   

3. Search your code:

   "Find code related to user authentication"
   

Explore the Help System

Ask your AI to run:

how_to_use()

To see an overview of all available capabilities.

What’s Next

We’re continuing to enhance Remembrances MCP with:

• More language support for code indexing
• Advanced code analysis features
• Performance optimizations for large codebases

Thank you to everyone who provided feedback and feature requests. Your input shapes the future of Remembrances MCP!

Found an issue? Have a feature request? Open an issue on GitHub!

• ←Previous
• Next→