Loading...
MCPAUDITOPEN SOURCE
TradeMemory Protocol
Persistent memory + audit trail for AI trading agents
The industry's first MCP-native memory layer for AI trading agents. 5 memory types (episodic, semantic, procedural, affective, prospective), outcome-weighted recall, and SHA-256 tamper-evident decision records. Open source, MIT licensed, 1,514 tests.
1,514Tests Passing
17MCP Tools
35+REST Endpoints
SHA-256Tamper Detection
THE CHALLENGE
AI trading agents start every session from zero
No memory of past trades. No context for decisions. No record of why a trade was taken. Every session is a blank slate, repeating the same mistakes.
THE SOLUTION
5-layer outcome-weighted memory architecture
Episodic Memory
Every trade stored with full decision context and power-law decay
Semantic Memory
Bayesian strategy beliefs auto-updated from trade outcomes
Affective State
EWMA confidence tracking + drawdown-linked risk appetite
Audit Trail
SHA-256 tamper-evident decision records
IN ACTION
TradeMemory Protocol
RESULTS
20
MCP tools exposed
100%
Decision audit coverage
1,519
Tests passing in CI
PythonFastAPISQLiteMCPClaude APISHA-256PostgreSQLpytest
Want something like this for your trading operation?
▸ Founder-led builds — limited availability