GammaShield
Autonomous AI Position Manager for DeFi Options
Created on 5th September 2026
•
GammaShield
Autonomous AI Position Manager for DeFi Options
The problem GammaShield solves
DeFi options are useful for hedging, but managing them is still difficult. A user has to watch spot prices, expiry, liquidity, dealer positioning, and the changing options book, then decide whether to buy protection, hold, close, or roll a position. That is a lot to monitor manually, especially when market conditions move quickly.
The problem: options protection is valuable, but interpreting the market and acting safely is too manual for most users.
GammaShield turns this into a guarded workflow, not an open-ended trading bot.
It reads the live Thetanuts OptionBook for BTC and ETH on Base, combines it with live spot-market data, and explains the market's dealer-gamma and liquidity conditions in a single dashboard. The user can review a real options quote and make a normal wallet-signed trade. For autonomous hedging, the user first creates a dedicated ERC-4337 policy account and signs a mandate with clear limits: asset, maximum loss, trade size, allowed actions, risk thresholds, persistence, cooldown, expiry, and pause/revoke controls.
The agent monitors the market, but it cannot spend outside those signed rules. Before any action, GammaShield rechecks the market, confirms that a fresh eligible quote exists, and applies the on-chain policy guard. This makes options protection easier to use while keeping the user in control of the money and the rules.
In practice, GammaShield helps users:
| What a user needs | How GammaShield helps |
|---|---|
| Understand a fast-moving options market | Translates the live Thetanuts book, spot prices, dealer gamma, liquidity, and expiry into an understandable risk view. |
| Review a potential options trade | Surfaces real Thetanuts OptionBook opportunities and fresh SDK-built quotes. |
| Automate cautiously | Lets the user set bounded hedging rules instead of handing a private key to a bot. |
| Stay in control | Supports pause and revoke controls before any later policy-account fill. |
| Learn safely | Provides a complete Base Sepolia policy-account flow before using Base mainnet. |
Challenges we ran into
| Challenge | What we did | Why it matters |
|---|---|---|
| Making autonomous execution safe instead of simulated | Replaced fixed application-level guards with owner-signed EIP-712 mandates and ERC-4337 policy accounts. | The policy account enforces signed limits, pause state, revocation, risk evidence, persistence, and cooldown—not a backend private key or hardcoded cap. |
| Separating a live quote from an executable order | Kept RFQ estimates advisory and enable trades only after a fresh SDK preview validates a listed order. | The agent cannot treat a price estimate, synthetic contract, or stale order as executable. |
| Testing without misrepresenting a testnet as Thetanuts mainnet | Built ShadowOptionBook on Base Sepolia for policy signing, funding, UserOperations, execution receipts, and positions. | Users can test the guarded flow without real funds, while the UI clearly distinguishes Shadow positions from Thetanuts positions. |
| Operating under real RPC and upstream limits | Narrowed and paged chain-log queries, added bounded retries and fallbacks, and cached short-lived snapshots. | The dashboard handles rate limits and upstream failures clearly instead of silently showing mock data. |
| Making AI useful without letting it override the user | Treated AI as a policy-drafting and second-opinion assistant; deterministic logic and the signed policy remain the final gate. | An unavailable or malformed AI response cannot bypass the user's limits or create a trade. |
Tracks Applied (3)
BEST PRODUCT BUILT ON THE THETANUTS SDK
Thetanuts Finance
AI × OPTIONS
Thetanuts Finance
AI For Society
Gonka Router
Technologies used