What is TradeMemory Protocol?

TradeMemory Protocol is a platform-agnostic AI trading memory system with Outcome-Weighted Memory (OWM) architecture via MCP. It stores trades, discovers patterns, generates strategy adjustments, and provides outcome-weighted recall. Works with MT5, Claude Desktop, Cursor, and any MCP client.

How does the memory architecture work?

Layer 1 records raw trade data (entry, exit, P&L, reasoning). Layer 2 runs a reflection engine to discover patterns (win rates by session, strategy, confidence). Layer 3 generates actionable strategy adjustments like disabling losing strategies or adjusting position sizes.

Does TradeMemory work with MT5?

Yes. TradeMemory includes mt5_sync.py that connects to MetaTrader 5 via the Python API, syncing closed trades automatically to the memory layer. It also supports any platform that can send trade data via the MCP protocol or REST API.

What is OWM (Outcome-Weighted Memory)?

OWM is a cognitive science-based recall system that scores past trades by outcome quality, context similarity, recency, confidence, and affective state. It applies established research (ACT-R, Kelly Criterion, Bayesian updating) to AI trading agents. Full specification: docs/OWM_FRAMEWORK.md

Open Source MCP Server

AI Trading Memory for MT5/Forex Traders

Your AI agent forgets every trade after each session. TradeMemory gives it persistent memory — store decisions, discover patterns, evolve strategy automatically via MCP.

Get Started View on GitHub

claude desktop — tradememory session

You: Store my XAUUSD trade: long 0.10 lots, +$1,175

✓ Trade stored: MT5-2350458751 | strategy=VolBreakout

You: Run a reflection on my last 73 trades

✓ Analyzing 73 trades across 4 strategies...

→ Pattern: IntradayMomentum BUY PF=2.11 return=+166% n=73

→ Pattern: MeanReversion SELL PF=0.79 return=−21% n=36

✓ L3 Adjustment: MeanReversion SELL → disable (consistent losses)

✓ L3 Adjustment: IntradayMomentum BUY → increase size 1.2x

Memory updated. 3 layers synced. 10,169 trades analyzed.

Core Features

Everything your trading agent needs to remember.

Six capabilities that turn a stateless AI into a learning trading partner.

▱

OWM Architecture

Outcome-Weighted Memory

5 cognitive memory types: Episodic, Semantic, Procedural, Affective, Prospective. Each trade scored by outcome quality, context similarity, recency, and confidence.

Score = Q × Sim × Rec × Conf × Aff

⚙

AI Reflection

Reflection Engine

Rule-based analysis finds patterns in your trade history — win rates by session, strategy, confidence level. Optional LLM layer for deeper insights.

reflection: london_session WR=73% ✓ strong
reflection: asia_session WR=38% ✗ weak

★

Pattern Discovery

Automatic Pattern Mining

L2 discovers which strategies, sessions, and conditions produce winners. L3 generates actionable adjustments: disable losers, size up winners.

pattern: IntradayMomentum BUY +47% n=73
pattern: MeanReversion SELL -21% n=36

📋

Daily Reports

Automated Daily Reflection

Schedule daily_reflection.py to run at market close. Summarizes P&L, flags anomalies, and updates memory layers automatically.

$ python daily_reflection.py
✓ Daily P&L: +$1,175.09 (1 trade)
✓ Patterns updated. Report saved.

📊

Behavioral Analysis

Bias Detection

Procedural memory tracks trading behaviors. Detects overtrading, revenge trading, and session-specific biases from your trade history.

overtrading: 3 trades in 12min → flag
revenge: loss→re-entry 2min → flag

⚗

Position Sizing

Kelly-from-Memory

Context-weighted Kelly Criterion using recalled win rates and payoff ratios. Quarter-Kelly default with risk appetite adjustment.

kelly_fraction: 0.12
recommended_size: 0.03 lots

Architecture

Dual Architecture

Legacy L1→L2→L3 pipeline plus OWM cognitive memory. Both work side by side.

LAYER 1 — RAW TRADES

Trade Storage

Every trade recorded with full context: instrument, direction, lots, entry/exit prices, P&L, strategy name, session, reasoning, and confidence score.

"instrument": "XAUUSD"
"direction": "long"
"pnl": +1175.09
"strategy": "VolBreakout"

LAYER 2 — PATTERNS

Pattern Discovery

Reflection engine analyzes L1 data to find statistically significant patterns. Groups by strategy, session, direction, and confidence level.

london_breakout: WR 73% n=41
asia_reversal: WR 38% n=12
high_confidence: WR 68% n=55

LAYER 3 — ADJUSTMENTS

Strategy Tuning

Auto-generates actionable adjustments from L2 patterns. Disable losing strategies, prefer winning sessions, adjust position sizes, restrict directions.

✓ asia_reversal: size × 0.5
✓ mean_reversion: disable SELL
✓ london_breakout: full size

OWM

OWM — COGNITIVE MEMORY

Outcome-Weighted Recall

5 memory types scored by outcome quality, context similarity, recency, confidence, and affective state. Based on ACT-R cognitive architecture.

Score = Q × Sim × Rec × Conf × Aff

Get Started

Install in 60 seconds.

Works with Claude Desktop, Claude Code, Cursor, Windsurf, and any MCP client.

Recommended

Claude Desktop

Add to your claude_desktop_config.json and restart Claude Desktop.

uvx tradememory-protocol click to copy

CLI

Claude Code

One command to add TradeMemory as an MCP server.

claude mcp add tradememory -- uvx tradememory-protocol click to copy

PyPI

pip install

Install from PyPI for custom integrations or development.

pip install tradememory-protocol click to copy

claude_desktop_config.json

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

MCP Tools

10 tools. Full cognitive memory pipeline.

Core tools for storage and analysis. OWM tools for cognitive recall, behavioral analysis, and position sizing. Your AI agent calls these tools via the Model Context Protocol.

store_trade

Record a trade with full context: instrument, direction, lots, prices, P&L, strategy, reasoning.

recall_trades

Query trade history with filters: date range, instrument, strategy, session, direction.

get_performance

Calculate win rate, profit factor, average P&L, max drawdown, and Sharpe ratio.

run_reflection

Analyze trade history, discover patterns (L2), and generate strategy adjustments (L3).

OWM Tools (v0.4.0)

remember_trade

Store trade as episodic memory, auto-update semantic/procedural/affective layers.

recall_memories

Outcome-weighted recall with score breakdown per component.

get_behavioral_analysis

Procedural memory bias detection: overtrading, revenge trading, disposition effect.

get_agent_state

Affective state: confidence, risk appetite, drawdown tracking, recommended action.

create_trading_plan

Prospective memory: conditional plans triggered by market conditions.

check_active_plans

Match active plans against current market context, auto-expire old plans.

Pricing

Coming Soon

Hosted API. Zero infrastructure.

Self-host free forever. Or let us handle the infrastructure — cloud API with guaranteed uptime, auto-scaling, and team dashboards.

Trader

$29/mo

For individual traders running 1–2 strategies on a single account.

1,000 API credits/month
1 trading account
Full L1–L3 memory pipeline
Daily reflection reports
7-day data retention
Community support

Coming Soon

Pay-as-you-go credits

Need more? Buy additional credits anytime. No commitment, no expiry. Every API call — store, recall, reflect — costs 1 credit.

Starter

1,000

$10

Growth

5,000

$40

Scale

25,000

$150

Get early access pricing

Join the waitlist. Early bird subscribers get 50% off the first 3 months.

No spam. We'll only email you when the hosted API launches.

✓ You're on the list! We'll notify you at launch.

AI Trading Memory for MT5/Forex Traders

Everything your trading agent needs to remember.

Outcome-Weighted Memory

Reflection Engine

Automatic Pattern Mining

Automated Daily Reflection

Bias Detection

Kelly-from-Memory

Dual Architecture

Trade Storage

Pattern Discovery

Strategy Tuning

Outcome-Weighted Recall

Install in 60 seconds.

Claude Desktop

Claude Code

pip install

10 tools. Full cognitive memory pipeline.

Hosted API. Zero infrastructure.

Pay-as-you-go credits

Get early access pricing

Stop trading without memory.