tradememory-protocol

mcp
SUMMARY

Decision audit trail + persistent memory for AI trading agents. Outcome-weighted recall, SHA-256 tamper detection, 17 MCP tools.

README.md

TradeMemory Protocol


TradeMemory Protocol gives AI trading agents two things they lack: a compliance-grade audit trail for every decision, and persistent memory that learns from outcomes.

Every AI trading tool executes trades. None of them record why. TradeMemory captures the full decision context — what conditions triggered the signal, which filters passed or blocked, the market indicators at that moment, risk state, and execution details. Every record is SHA-256 hashed for tamper detection. And across sessions, your agent remembers what worked, discovers patterns, and adjusts strategy automatically — using a three-layer architecture inspired by ACT-R cognitive science.

When to use this: You're building an AI agent that trades forex, crypto, or equities via MT5, Binance, Alpaca, or any platform — and you need it to (1) prove why it made each decision, and (2) remember what worked across sessions.

How it works

  1. Audit — Every decision is recorded with full context: conditions evaluated, filters checked, indicators at that moment, risk state. SHA-256 hashed at creation for tamper detection.
  2. Store — Your agent records trades with context (strategy, confidence, market regime) via MCP tools
  3. Recall — Before the next trade, the agent retrieves similar past trades weighted by outcome (Outcome-Weighted Memory)
  4. Evolve — The Evolution Engine discovers patterns across trades and generates new strategy hypotheses, validated with Deflated Sharpe Ratio

When to use TradeMemory vs alternatives

TradeMemory Raw Mem0/Qdrant LangChain Memory Custom SQLite
Decision audit trail ✅ SHA-256 + TDR ❌ None ❌ None ❌ DIY
Trade-specific schema ✅ L1→L2→L3 pipeline ❌ Generic vectors ❌ Chat-oriented ❌ DIY everything
Outcome weighting ✅ Kelly + ACT-R ❌ Cosine only ❌ Recency only ❌ Manual
Strategy evolution ✅ Built-in engine ❌ Not included ❌ Not included ❌ Not included
MCP native ✅ 17 tools ❌ Custom wrapper ❌ Custom wrapper ❌ Custom wrapper
Statistical validation ✅ DSR + walk-forward ❌ None ❌ None ❌ None

News

  • [2026-03] Decision Audit Trail — Trading Decision Records (TDR) with SHA-256 tamper detection, 4 audit REST endpoints, 2 MCP tools, JSONL decision context ingestion
  • [2026-03] Onboarding CLItradememory setup wizard, doctor health check, 8-platform config generator
  • [2026-03] v0.5.0 — Evolution Engine + OWM 5 memory types. Release Notes
  • [2026-03] Statistical Validation — Strategy E passes P100% random baseline, Sharpe 3.24 walk-forward
  • [2026-02] v0.4.0 — OWM Framework, 15 MCP tools, Smithery + Glama listed

Architecture

Architecture

Three-Layer Memory

L1 L2 L3 Memory Pipeline

Decision Audit Trail

Every trading decision your agent makes — including decisions not to trade — is recorded as a Trading Decision Record (TDR). Each record captures the full reasoning chain and is SHA-256 hashed at creation for tamper detection.

Here is a real decision event from a XAUUSD trading system. The AI agent detected a SHORT breakout signal but the sell_allowed filter blocked execution:

{
  "ts": "2026-03-26 07:55:00",
  "strategy": "VolBreakout",
  "decision": "FILTERED",
  "signal_triggered": true,
  "signal_direction": "SHORT",
  "conditions_json": {
    "conditions": [
      {"name": "breakout_high", "passed": false, "current_value": 4462.58, "threshold": 4569.75},
      {"name": "breakout_low", "passed": true, "current_value": 4462.58, "threshold": 4463.11}
    ]
  },
  "filters_json": {
    "filters": [
      {"name": "spread_gate", "passed": true, "current_value": 12.0},
      {"name": "sell_allowed", "passed": false, "blocked": true},
      {"name": "account_risk", "passed": true},
      {"name": "regime_gate", "passed": true}
    ]
  },
  "indicators_json": {
    "atr_d1": 171.16, "atr_m5": 8.53,
    "asia_high": 4544.08, "asia_low": 4488.78, "asia_range": 55.30
  },
  "regime": "TRENDING",
  "consec_losses": 0,
  "cooldown_active": false,
  "risk_daily_pct": 0.0
}

A regulator or risk manager can read this and immediately understand: the agent saw a valid breakout, but policy blocked the SHORT direction. No guessing, no black box.

Audit API

# Get full decision record
GET /audit/decision-record/{trade_id}

# Verify record hasn't been tampered with
GET /audit/verify/{trade_id}
# → {"verified": true, "stored_hash": "a3f8c9...", "computed_hash": "a3f8c9...", "match": true}

# Bulk export for regulatory submission
GET /audit/export?strategy=VolBreakout&start=2026-03-01&end=2026-03-31&format=jsonl

Regulatory Alignment

Regulation Requirement TradeMemory Coverage
MiFID II Article 17 Record every algorithmic trading decision factor Full decision chain: conditions, filters, indicators, execution
EU AI Act Article 14 Human oversight of high-risk AI systems Explainable reasoning + memory context for every decision
EU AI Act Logging Systematic logging of every AI action and decision path Automatic per-decision TDR with structured JSON
ESMA 2026 Briefing Algorithms must be distinguishable, testable, identifiable agent_id + model_version + strategy per record

Quick Start

pip install tradememory-protocol

Try the demo (no API key needed):

tradememory demo

Add to claude_desktop_config.json:

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

Then tell Claude: "Record my BTCUSDT long at 71,000 — momentum breakout, high confidence."

Claude Code / Cursor / Docker
# Claude Code
claude mcp add tradememory -- uvx tradememory-protocol

# From source
git clone https://github.com/mnemox-ai/tradememory-protocol.git
cd tradememory-protocol && pip install -e . && python -m tradememory

# Docker
docker compose up -d

Setup & Configuration

First-time guided setup:

tradememory setup

This walks you through:

  1. Terms acceptance — trading disclaimer and data storage policy
  2. Platform detection — auto-detects Claude Desktop, Claude Code, Cursor, Windsurf, Cline
  3. Config generation — prints the exact JSON snippet for your platform
  4. Health check — verifies database, MCP tools, and core functionality

Platform Configs

Generate config for any supported platform:

tradememory config              # interactive menu
tradememory config claude_code  # direct: auto-installs via CLI
tradememory config cursor       # prints .cursor/mcp.json snippet
tradememory config windsurf     # prints Windsurf config
tradememory config raw_json     # generic MCP JSON

Supported: Claude Desktop · Claude Code · Cursor · Windsurf · Cline · Smithery · Docker

Health Check

tradememory doctor        # core checks (~3s)
tradememory doctor --full # + REST API, MT5, Anthropic API

MCP Tools (17)

Category Tools
Core Memory store_trade_memory · recall_similar_trades · get_strategy_performance · get_trade_reflection
OWM Cognitive remember_trade · recall_memories · get_behavioral_analysis · get_agent_state · create_trading_plan · check_active_plans
Evolution evolution_fetch_market_data · evolution_discover_patterns · evolution_run_backtest · evolution_evolve_strategy · evolution_get_log
Audit export_audit_trail · verify_audit_hash
REST API (35+ endpoints)

Trade recording, outcome logging, history, reflections, risk constraints, MT5 sync, OWM, evolution, decision audit trail, integrity verification.

Full reference: docs/API.md

OWM — Outcome-Weighted Memory

OWM 5 Factors

Full theoretical foundation: OWM Framework

Evolution Engine

Evolution Engine

Methodology & data: Research Log

Security

  • TradeMemory never touches API keys. It does not execute trades, move funds, or access wallets.
  • Read and record only. The agent calls TradeMemory after making a decision, passing the context. TradeMemory stores it.
  • No external network calls. The server runs locally. No data is sent to third parties.
  • SHA-256 tamper detection. Every record is hashed at creation time. Verify integrity at any point with /audit/verify.
  • 1,233 tests passing. Full test suite with CI.

Documentation

Doc Description
Architecture System design & layer separation
OWM Framework Full theoretical foundation
Tutorial Install → first trade → memory recall
API Reference All REST endpoints
MT5 Setup MetaTrader 5 integration
Research Log 11 evolution experiments
Roadmap Development roadmap
中文版 Traditional Chinese

Contributing

See Contributing Guide · Security Policy

Star History

MIT — see LICENSE. For educational/research purposes only. Not financial advice.

Built by Mnemox

Reviews (0)

No results found