Tradememory Protocol README

Project status (August 2026): Feature-complete, in maintenance mode — bug and security reports are still reviewed; no new features or hosted service are planned. For paid work, see Trading Record Analysis.

Your trading AI has amnesia. And regulators are starting to notice.

It makes the same mistakes every session. It can't explain why it traded. It forgets everything when the context window ends. Meanwhile, MiFID II is raising the bar for algorithmic decision documentation (Article 17). The EU AI Act demands systematic logging of AI actions (Article 14). Your competitors' agents are learning from every trade.

The AI trading stack is missing a layer. Every MCP server handles execution — placing orders, fetching prices, reading charts. None handle memory.

Your agent can buy 100 shares of AAPL but can't answer: "What happened last time I bought AAPL in this condition?"

TradeMemory is the memory layer. One pip install, and your AI agent remembers every trade, every outcome, every mistake — with a SHA-256 tamper-evident audit trail.

Used in production by traders running pre-flight checklists before every position, and by EA systems logging thousands of decisions daily.

What it does

• Before trading: ask your memory — what happened last time in this market condition? How did it end?
• After trading: one call records everything — five memory layers update automatically
• Safety rails: confidence tracking, drawdown alerts, losing streak detection — the system tells you when to stop

Works with any market (stocks, forex, crypto, futures), any broker, any AI platform. TradeMemory doesn't execute trades or touch your money — it only records and recalls.

See the interface

tradememory-dashboard.onrender.com — the dashboard running on an illustrative demo dataset. Nothing to install.

It is an interface preview, not a track record: the trades are synthetic and every figure on it is labelled as such. For what the memory layer actually does in a terminal, pip install tradememory-protocol && tradememory demo --fast replays 30 trades and shows the recall and parameter adjustment it derives from them.

Quick Start

pip install tradememory-protocol

Add to Claude Desktop (claude_desktop_config.json):

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

Then tell Claude: "Record my AAPL long at $195 — earnings beat, institutional buying, high confidence."

# 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

Full walkthrough: Getting Started (Trader Track + Developer Track)

Who uses TradeMemory

How it works

1. Recall — Before trading, retrieve past trades weighted by outcome quality, context similarity, recency, confidence, and emotional state (OWM Framework)
2. Record — After trading, one call to remember_trade writes to five memory layers: episodic, semantic, procedural, affective, and trade records
3. Reflect — Daily/weekly/monthly reviews detect behavioral drift, strategy decay, and trading mistakes
4. Audit — Every decision is SHA-256 hashed at creation. Export anytime for review or regulatory submission

MCP Tools

Trading Record Analysis

TradeMemory itself is free and self-hosted. What the maintainer offers as a paid service is statistical analysis of your own trading records: export your MT4/MT5 history and get a descriptive-statistics report — where your losses concentrate, how your position sizing changes after losses, forced-liquidation structure, and the actual risk you took per trade — followed by a walkthrough call.

Descriptive statistics of past trades only: no trade signals, no investment advice, no performance promises. Your files are deleted after delivery.

dev@mnemox.ai | Book a call

Enterprise & Compliance

Every trading decision your agent makes — including decisions not to trade — is recorded as a Trading Decision Record (TDR). Per-record SHA-256 content hashes are linked into a forward-chained audit ledger; every UTC day is summarised by a Merkle root which itself chains across days. Tampering with any historical record invalidates every subsequent link.

# Verify a single record hasn't been tampered with
verify_audit_hash(trade_id="MT5-7047640363")
# → {"verified": true, "chain_entry": {"sequence_num": 42, ...}}

# Walk the entire chain (or a slice) end-to-end
verify_audit_chain(from_seq=1, to_seq=None)
# → {"verified": true, "checked_count": 1284, "first_break_at": null}

# Daily Merkle root — single 32-byte anchor over every TDR for that day
get_daily_root(date="2026-05-14")
# → {"verified": true, "root_hash": "a05544...", "record_count": 18}

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

See LIMITATIONS.md for the full audit-chain maturity statement, including what's not in v0.5.2 yet (TSA timestamping, external anchoring, zkML proof of inference).

Need a custom deployment for your fund? → dev@mnemox.ai

Security

• Never touches API keys. TradeMemory does not execute trades, move funds, or access wallets.
• Read and record only. Your agent passes decision context to TradeMemory. It stores it. That's it.
• Local-first. The only outbound call is RFC 3161 trusted timestamping of daily audit roots — a 32-byte hash, no trade data (on by default; disable with TRADEMEMORY_TSA=off). Nothing else leaves your machine.
• SHA-256 chained audit ledger. Every record is hashed at creation and linked to the previous record. Daily Merkle roots anchor the chain. Verify integrity at the record, slice, or day level. Tampering is detectable at every level; external anchoring (TSA by default) is on the roadmap.
• 1,400+ tests passing. Full test suite with CI.

Research Status

TradeMemory's OWM framework is grounded in cognitive science (Tulving 1972) and reinforcement learning (Schaul et al. 2015). Current status:

• OWM five-factor scoring: implemented, tested (1,400+ tests)
• Statistical validation: DSR, MBL implemented (Bailey-de Prado 2014)
• Audit trail: SHA-256 tamper-evident TDR
• Evolution engine: research phase (strategy generation works, statistical gate pass rate under optimization)
• Hybrid recall: OWM-only mode active, vector fusion available when embeddings configured
• Empirical validation: ongoing (n=40 trades, target n>=100 for statistical significance)

Documentation

Contributing

See Contributing Guide · Security Policy

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