AI Agents + ZK Privacy on Uniswap
Agents that trade on your behalf, coordinate with each other, and keep your strategy private — powered by a primitive that didn't exist before: ZK Private Delegation.
🤖
Agents That Trade
Natural-language intents parsed in <1ms, deterministically, client-side. The agent executes swaps, LP positions, and rebalances on Uniswap V3 — only when confidence is 100%.
🤝
Coordinate With Other Agents
Swarm agents share market intelligence, reach consensus on opportunities, and coordinate execution across Uniswap V3 and Balancer — without exposing individual strategies.
🔐
A New Primitive: ZK Private Delegation
Delegate trading authority to agents with a Groth16 ZK proof verified on-chain. The agent can execute, but your intent, limits, and strategy stay hidden. Deployed on Sepolia.
Circom Circuit
·
Groth16 Prover
·
Solidity Verifier
·
Uniswap V3 Router
·
Sepolia Testnet
·
Zero API Cost
The Problem We Solve
When you delegate to an agent today, your strategy is public. We fix that with ZK proofs.
TRADING INTENT
"Manage my liquidity across Uniswap V3, Curve, and Balancer. Optimize for highest APY with max 50 ETH daily limit. Focus on blue-chip pools only."
Public Delegation (Current Way)
Everyone sees your delegation details
DELEGATION DETAILS (PUBLIC)
User:
0x742d...8b3a
Delegate:
0x9f2e...1c4d
Daily Limit:
50 ETH
Protocols:
Uniswap, Curve, Balancer
Strategy:
Highest APY, Blue-chip only
RISKS EXPOSED
FRONT-RUNNING
Traders see your 50 ETH limit
STRATEGY COPYING
Competitors copy your pool selection
PRICE IMPACT
Market moves against your trades
PRIVACY LOSS
Trading patterns fully exposed
FINANCIAL IMPACT
Expected APY:
12.5%
Actual APY:
8.2%
Loss to Front-running:
-4.3%
ZK Private Delegation (Our Way)
Only prove authorization, hide all details
DELEGATION DETAILS (PRIVATE)
User:
0x742d...8b3a
Delegate:
0x9f2e...1c4d
Daily Limit:
🔒 PRIVATE
Protocols:
🔒 PRIVATE
Strategy:
🔒 PRIVATE
ZERO-KNOWLEDGE PROOF
A = g^secret mod p
T = g^random mod p
challenge = Hash(A || T || context)
s = random + secret * challenge mod p
✓ Valid authorization proven
✓ All details kept private
✓ All details kept private
PRIVACY BENEFITS
NO FRONT-RUNNING
Limits are hidden
STRATEGY PROTECTION
Pool selection stays secret
BETTER EXECUTION
No price impact from leaks
FINANCIAL PRIVACY
Patterns stay confidential
FINANCIAL IMPACT
Expected APY:
12.5%
Actual APY:
12.3%
Privacy Premium:
+4.1%
Why Not Just Use an LLM?
LLMs are better at understanding language. That's not the point. The question is: would you trust a probabilistic model to decide how much ETH to deploy?
The Problem With LLMs for DeFi Execution
LLMs are great at conversation. But when you say "manage my liquidity across Uniswap V3 with max 25 ETH daily limit", that intent needs to become executable parameters — the exact protocol, the exact amount, the exact constraints. This is structured data extraction, not conversation. For this job, you need:
🎯
Determinism
Same input = same output, every time. Ask an LLM the same question twice and you can get different parameter extractions. When you're moving 25 ETH, "close enough" isn't good enough.
⚡
Speed
LLM API round-trips take 2-10 seconds. MEV bots operate in milliseconds. By the time an LLM parses your intent, a front-runner has already seen and acted on your transaction.
🔒
Privacy
Sending "50 ETH daily limit on Uniswap V3 blue-chip pools" to OpenAI means your strategy is logged on their servers. For DeFi, your intent IS your alpha.
🛡️
Zero Failure Modes
No API key expiration. No rate limits. No OpenAI outage at 2am when your auto-rebalance triggers. No content filter refusing to help with "risky" DeFi strategies.
LLM (ChatGPT, Claude, etc.)
Better at understanding language
WHAT LLMs DO WELL
✓
Understands ambiguous, conversational intents
✓
Can explain strategies and answer follow-ups
✓
Handles any language, any domain
✓
Gets smarter with each model update
WHY THEY'RE WRONG FOR EXECUTION
✗
Non-deterministic — same prompt, different parameters each time
✗
Too slow — 2-10s round-trip, MEV bots act in ms
✗
Leaks your strategy — intent sent to 3rd-party API servers
✗
Can fail — rate limits, outages, content filters on "risky" DeFi
✗
Hallucinations — can invent parameters that weren't in the intent
1Hundo Intent Engine
Purpose-built for DeFi parameter extraction
BUILT FOR DEFI EXECUTION
✓
Deterministic — same input always produces same executable params
✓
Sub-millisecond — parses before MEV bots can react
✓
DeFi-native — knows token pairs, fee tiers, pool types, risk profiles
✓
Never hallucinates — only extracts what's actually in the intent
✓
Validates — warns about missing params, suggests improvements
HONEST LIMITATIONS
~
Only handles DeFi intents — not a general chatbot
~
New protocols need to be added to the knowledge base
~
Can't explain "why" — it parses, it doesn't reason
THE 100% RULE
1Hundo will not execute unless confidence is 100%. If anything in your intent is unclear — the action, the protocol, the amount — we refuse the transaction and tell you exactly what's missing. The burden of understanding is on us, not you. An LLM will happily hallucinate the missing parameters and execute anyway.
Try It Yourself — Live DeFi Intent Parsing
Type any DeFi intent and see 1Hundo's deep parsing in real-time
Live Comparison
See the difference in real-time
Public Transaction
TRANSACTION DETAILS
From: 0x742d...8b3a
To: 0x9f2e...1c4d
Amount: 25 ETH
Protocol: Uniswap V3
Delegation: 50 ETH daily limit ✓
⚠️ Front-runners can see this coming!
Private Transaction
TRANSACTION DETAILS
From: 0x742d...8b3a
To: 0x9f2e...1c4d
Amount: 25 ETH
Protocol: Uniswap V3
Authorization: ZK Proof ✓
✓ No delegation details exposed!