Phase 0 beta is live. Trade perps with AI agents — stay invisible.

How GitSync Works

A five-layer architecture that keeps your trades private from start to finish.

GitSync replaces the traditional exchange model — where you broadcast orders into a public book — with a private, AI-driven pipeline. Every trade flows through five layers, each adding a specific privacy or execution guarantee.

The Five Layers

1
Policy Interface

You don't place orders. You define a policy — a natural-language mandate with hard risk caps (max drawdown, max leverage, instrument whitelist, time horizon). Your policy is signed and encrypted before ever leaving your browser.

2
Private AI Agent Runtime

Each policy spawns an isolated AI agent. The agent ingests real-time market data, runs inference, and emits two outputs at every tick: a proposed intent (the trade) and an explanation (plain-language rationale shown to you before execution).

3
Confidential Intent Matcher

Your intent is encrypted under a threshold committee key. It enters a batch with other encrypted intents. At epoch close (~500ms), the committee decrypts and matches all intents at a single uniform clearing price. No individual order is ever published in cleartext.

4
Shielded Position State

Positions are stored as cryptographic commitments on-chain — not plaintext records. Your direction, size, leverage, and liquidation price are hidden from everyone except you. State transitions are verified with zero-knowledge proofs.

5
Base L2 Settlement

All settlement happens on Base, Ethereum's most cost-efficient L2. The aggregate batch, ZK proofs, and vault updates are all written to Base, giving you the finality and auditability of an EVM rollup — at a fraction of mainnet cost.

End-to-end flow: Set policy → Agent decides → Intent encrypted → Batch matched → Position committed → Settled on Base. At no point is your strategy, position, or intent visible to extractors, block builders, or counterparties.

Last updated just now