In Memoria

What it does

In Memoria is an MCP server that solves "session amnesia" for AI coding assistants. It learns your codebase's architectural patterns, naming conventions, and structural preferences, storing them in a local database so that AI tools like Claude and Copilot can maintain continuous context without needing to re-analyze the project every session.

Tools

• analyze_codebase: Analyzes files or directories for concepts, patterns, and complexity.
• search_codebase: Multi-mode semantic, text, and pattern-based search.
• learn_codebase_intelligence: Deep learning to extract architectural patterns.
• get_project_blueprint: Provides instant project context (tech stack, entry points) in <200 tokens.
• predict_coding_approach: Implementation guidance with smart file routing.
• auto_learn_if_needed: Smart auto-learning with staleness detection.
• get_semantic_insights: Queries learned concepts and relationships.

Installation

{
  "mcpServers": {
    "in-memoria": {
      "command": "npx",
      "args": ["in-memoria", "server"]
    }
  }
}

Supported hosts

• Claude Desktop
• Claude Code CLI
• GitHub Copilot