mnemox-ai/tradememory-protocol

Decision audit trail + persistent memory for AI trading agents. Outcome-weighted recall, tamper-evident SHA-256 chain with RFC 3161 anchoring, 20 MCP tools.

What it solves

TradeMemory provides a dedicated memory layer for AI trading agents to prevent them from repeating mistakes and to meet regulatory requirements. It solves the problem of "amnesia" in AI agents—where they forget past trades and the reasoning behind them once the context window ends—by providing a persistent, structured record of every decision.

How it works

TradeMemory acts as a non-executing recording and recall system that integrates with AI platforms via the Model Context Protocol (MCP). It uses the Outcome-Weighted Memory (OWM) framework to store and retrieve trades across five memory layers (episodic, semantic, procedural, affective, and trade records).

Key processes include:

  • Recall: Retrieving past trades based on outcome quality, context similarity, and recency before a new trade is placed.
  • Record: Saving the details of a trade and its outcome after execution.
  • Reflect: Analyzing behavioral drift and strategy decay over time.
  • Audit: Creating a SHA-256 tamper-evident audit trail where every decision is hashed and linked into a forward-chained ledger with daily Merkle roots.

Who it’s for

  • AI Trading Developers: Those building agents that need long-term memory and behavioral analysis.
  • Quantitative Traders: Users running pre-flight checklists to maintain discipline.
  • Compliance Teams: Funds and institutions needing a tamper-resistant audit trail to satisfy regulations like MiFID II and the EU AI Act.

Highlights

  • MCP Integration: Ready-to-use tools for Claude Desktop, Cursor, and other MCP-compatible agents.
  • Tamper-Evident Audit Trail: Uses SHA-256 hashing and Merkle roots to ensure records cannot be altered without detection.
  • Outcome-Weighted Memory: A cognitive-science-based approach to recall trades based on their success or failure.
  • Safety Rails: Built-in tracking for confidence, drawdowns, and losing streaks to alert the agent when to stop trading.

Related

  • Project
    agentic-box/memora

    An MCP memory layer for AI agents that provides persistent, structured storage with semantic retrieval and knowledge graph visualization.

  • Project
    Goldentrii/AgentRecall-X

    A long-term memory system for AI agents that tracks and measures whether human corrections actually change agent behavior across sessions.

  • Project
    mnemon-dev/mnemon

    A persistent, cross-session memory system for AI agents that uses the host LLM as a supervisor to manage a four-graph knowledge store.

  • Project
    CaviraOSS/OpenMemory

    A cognitive memory engine for AI agents that provides long-term, self-hosted memory with temporal reasoning and multi-sector categorization instead of simple RAG.

  • Project
    MemoriLabs/Memori

    A persistent, structured memory layer for AI agents and LLMs that automatically captures and recalls user preferences and agent actions to reduce token costs and maintain context across sessions.