SymbolicMemoryMCP: An Enhanced Symbolic Memory Solution for AI Systems

SymbolicMemoryMCP is a cutting-edge plugin designed to deliver explicit and deterministic symbolic memory for AI systems using the Model Context Protocol (MCP). This project provides a minimal MCP server that enables language models (LLMs) and AI agents to efficiently store and retrieve curated “ground truth” using stable symbols and optional aliases, moving beyond traditional probabilistic recall methods.

Key Features

• Deterministic Recall: SymbolicMemoryMCP ensures accurate retrieval, eliminating guesswork.
• Stable References: Symbols remain constant over time, providing reliability.
• Curated Ground Truth: Allows for high-quality information storage that is human-verifiable and low in noise.

Components of the Repository

• server.py: A FastAPI-based server implementing the MCP JSON-RPC protocol.
• client.py: A command-line interface for manual save and get operations.
• tests_smoke.py: End-to-end smoke tests for the MCP functionality.
• MCP2genericLLM.py: A reference bridge for LLM integration (tested with Ollama).

Understanding the Value of SymbolicMemoryMCP

Current AI systems typically remember data through chat histories, vector-based memory, and dynamic prompt states—in each case, data retrieval can be imprecise and time-consuming. SymbolicMemoryMCP addresses these challenges by adding a layer of deterministic memory that can be effectively utilized alongside existing approaches:

• Use vector memory for broad, fuzzy memory traces.
• Leverage SymbolicMemoryMCP for concise, curated definitions and invariants.

Symbol Model

A symbol serves as a stable and human-readable key, promoting consistency in reference keys through a recommended naming convention:

• Format: Uppercase segments separated by dots (e.g., DOMAIN.SUBDOMAIN.NAME).
• Optional aliases enable natural language-friendly alternatives for the same symbol.

Implemented MCP Methods

Saving Data: Use the tools/call method with sm.texts.save to store information.

{
  "symbol": "HGI.DEF",
  "text": "Hybrid General Intelligence = AI + human symbiosis",
  "cat": "ai",
  "subcat": "concepts.intelligence",
  "aliases": ["hgi", "hybrid intelligence"]
}

Retrieving Data: Use the resources/read method by referring to symbols or aliases:

• For example: resource://sm/v1/texts/HGI.DEF or resource://sm/v1/texts/hybrid intelligence.

Application Scenarios

For optimal functionality:

• Call sm_get for canonical definitions when correctness is essential.
• Utilize sm_save when the user provides explicit definitions to store.

Architecture Overview

The architecture integrates smoothly with current agent stacks, offering a reliable symbolic memory interface residing above an SQLite-backed storage layer, optimizing the accessibility and management of data within AI systems.

Future Roadmap

Future updates will include enhancements such as prefix search, explicit alias management, and versioning capabilities to accommodate dynamic semantic environments while maintaining a focus on a compact, deterministic core system.