Live on Base Sepolia

On-chain guardrails
for AI agent trading.

Prompt injection can confuse a model. It cannot bypass a smart contract. MoltFi puts your agent's trading limits on-chain — no bug, no bad prompt, no exploit can override them.

I'm an AI agent
curl -s https://moltfi-production.up.railway.app/api/skill
I'm a human
Connect your wallet above to create a vault and set trading policies. Then give your agent access — paste this into its chat:
Read this skill file and follow the instructions to manage my DeFi vault: curl -s https://moltfi-production.up.railway.app/api/skill

How It Works

You set the rules. Your agent trades within them. Smart contracts enforce every limit.

Your Agent
"Swap 0.1 WETH to USDC"
→
MoltFi Vault
Interprets request
→
AgentPolicy
Checks limits on-chain
→
Uniswap V3
Executes the swap
🔐

Create a vault

Connect your wallet, deposit funds, set guardrails. Limits are written to a smart contract — not your agent's config file.

🤖

Agent gets an API key

Your agent reads a skill file and registers. One API key, scoped to your vault. No private keys, no direct chain access.

💬

Trade in plain English

"Swap 0.01 WETH to USDC." Venice AI interprets. The smart contract checks limits. If it's within bounds, Uniswap executes.

✅

Verified on-chain

Every trade has a Basescan link. Over your limits? Transaction reverts automatically. No funds move. No exceptions.

You Set the Policies

Your wallet owns the vault. You configure exactly what your agent is allowed to do — enforced on-chain.

Max per trade

Set a per-trade limit. Every swap is checked on-chain before execution.

Daily spending cap

Set a daily volume cap. Tracked cumulatively, resets every 24 hours.

Token allowlist

Choose which tokens your agent can trade. Everything else is blocked.

Instant freeze

Revoke all trading with one transaction. Takes effect immediately.

Works With Any Agent

If your agent can make HTTP calls, it can use MoltFi. No SDK, no blockchain dependency.

✓ OpenClaw, ChatGPT, Claude — any agent
✓ Plain English requests — no ABI encoding
✓ API key auth — register once, trade anytime
✓ Your agent never touches private keys
# Your agent registers (once)
POST /api/agent/register
{"humanWallet": "0x..."}
# Returns: API key + vault address
# Your agent trades (anytime)
POST /api/agent
Authorization: Bearer mf_...
{"message": "swap 0.001 WETH to USDC"}
# Smart contract enforces your limits

Deployed on Base Sepolia

Real contracts, real swaps, verified on Basescan.

AgentPolicy →AgentGuardRouter →VaultFactory →

Built With

Real infrastructure. Real contracts. Real swaps.

⚡

OpenClaw

Agent runtime and skill harness. Kyro (AI co-builder) runs on OpenClaw — reading skill files, making API calls, managing vault operations, and coordinating the entire agent workflow autonomously.

🔵

Base

Deployed on both Base Sepolia and Base mainnet. AgentPolicy, AgentGuardRouter, and VaultFactory — live on-chain with real transactions.

🦄

Uniswap V3

Every swap executes through Uniswap V3 SwapRouter02 via our AgentGuardRouter. Real liquidity, real trades, real guardrail enforcement.

🔒

Venice AI

Private inference with zero data retention. Processes plain English trade requests into contract calls. Your agent's reasoning never gets stored.

💧

Lido

Live wstETH in our Base mainnet vault, earning ~2.9% APR. Swapped through our guardrail-enforced router. Real yield, real on-chain proof.