# mnemox-ai/tradememory-protocol [Health: Active]

**Category:** 💰 Finance & Fintech  
**Repository:** https://github.com/mnemox-ai/tradememory-protocol  
**GitHub Stars:** 1417  
**Views:** 4  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/mnemox-ai-tradememory-protocol

## Description
Glama 🐍 🏠 - Structured 3-layer memory system (trades → patterns → strategy) for AI trading agents. Supports MT5, Binance, and Alpaca.

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "tradememory-protocol": {
    "command": "uvx",
    "args": ["tradememory-protocol"]
  }
}
```

## Documentation

## What mnemox-ai/tradememory-protocol MCP server does

The mnemox-ai/tradememory-protocol MCP server provides a memory and governance layer for trading agents. It records trade details, outcomes, reasoning, market conditions, and related state so an agent can retrieve relevant prior decisions instead of relying only on its current context window.

The system is designed for stocks, forex, crypto, futures, and other markets. The repository describes use with MT5, Binance, and Alpaca, while also stating that the server does not place orders or manage funds. Execution remains the responsibility of a separate broker or trading system.

Trade records update episodic, semantic, procedural, affective, and trade-oriented memory. The server also supports trading plans, strategy performance, reflections, behavioral analysis, confidence tracking, drawdown and losing-streak checks, and pre-trade legitimacy decisions.

## How it works

A typical workflow has four stages. Before a trade, `recall_memories` retrieves prior experiences using factors including outcome quality, context similarity, recency, confidence, and emotional state. The agent can then use `check_trade_legitimacy` as a five-factor pre-trade gate that returns a full, reduced, or skip decision.

Afterward, `remember_trade` records the decision and updates the memory layers. Scheduled reviews can use reflection and behavioral tools to identify trading mistakes, behavioral drift, or strategy decay. Planning tools support prospective decisions with conditional triggers.

Audit records receive SHA-256 content hashes when created. The repository also describes a forward-linked audit chain and daily Merkle roots, with tools for checking individual hashes, validating chain ranges, retrieving daily roots, and exporting records for review.

## Setup and configuration

The mnemox-ai/tradememory-protocol MCP server can be installed from PyPI:

```bash
pip install tradememory-protocol
```

For a local MCP client, the documented launch command is:

```json
{
  "mcpServers": {
    "tradememory": {
      "command": "uvx",
      "args": ["tradememory-protocol"]
    }
  }
}
```

The README specifically shows configuration for Claude Desktop and a corresponding `claude mcp add` command for Claude Code. It also documents running from a source checkout with editable installation, plus a Docker Compose option. The project is self-hosted; no hosted service is planned according to its stated maintenance status.

## Tools and capabilities

The documented MCP tool groups include:

- Memory: `remember_trade`, `recall_memories`, `get_strategy_performance`, and `get_trade_reflection`
- State and behavior: `get_agent_state` and `get_behavioral_analysis`
- Planning: `create_trading_plan` and `check_active_plans`
- Risk and governance: `check_trade_legitimacy`, `validate_strategy`, and `compute_dqs`
- Evolution: market-data retrieval, pattern discovery, backtesting, strategy evolution, and evolution logs
- Audit: `export_audit_trail`, `verify_audit_hash`, `verify_audit_chain`, and `get_daily_root`

The repository also describes a REST API with more than 35 endpoints covering trade records, reflections, risk, MT5 synchronization, OWM functions, evolution, and audit operations.

## Limitations and notes

The mnemox-ai/tradememory-protocol MCP server is in feature-complete maintenance mode. The README says bug and security reports remain under review, but no new features or hosted service are planned. It records and analyzes trading decisions; it is not an execution server, investment-advice system, or source of trade signals.

The repository’s audit documentation includes maturity limitations for some planned audit features, including TSA timestamping. Review `LIMITATIONS.md` before treating the audit output as sufficient for a particular regulatory process. The paid offering described by the maintainer is separate statistical analysis of user-provided trading records, not a required component of the self-hosted server.

_Full upstream README: https://allmcps.com/mcp/mnemox-ai-tradememory-protocol/readme_