Problem
AI agents already handle onchain payments alone: the x402 protocol went from 188 millions of accumulated transactions. But the agent has no way of knowing if the destination contract is legitimate, if the vault is liquid, or if the transaction will undergo a frontrun or sandwich. In 2025, crypto hacks, and scams cost more than US$ 4 billions (PeckShield). The existing security check was done for humans, in wallets and manual audits. The agent who pays alone doesn't have that layer. The Hedwig agent is the opsec service that the payment agent consults via API or MCP before signing any transaction.
Solution
Hedwig runs as an MCP server and as a REST API. Before executing a payment, the agent submits the destination address and transaction details. We respond in milliseconds with a verdict and the reasons: verified or malicious contract, real vault liquidity, address history, risk of manipulation in Mempool. Behind it is its own indexing of contracts, malicious address bases, and transaction simulation. The agent receives a simple answer: proceed, adjust, or abort.
Business model
Charge per use: the customer pays for verification via API/MCP, with a unit price low enough to fit an x402 payment. High-volume platforms migrate to the monthly plan with a guaranteed minimum. Target ticket: US$ 50 the US$ 200 per month for individual developers, US$ 1.000+ per month for agent platforms.
Market
Two markets intersect. Crypto security: hacks and scams cost more than US$ 4 billions in 2025 (PeckShield), and security is a mandatory expense for those who operate on-chain. Agent payments: the x402 already surpasses 188 millions of transactions and US$ 42 millions liquidated (set/2026), with projections of US$ 3 the 5 trillions in trade by agents up to 2030 (Galaxy Research). We charge per verification, so every agent transaction is a potential call. Billing scales along with the protocol.
Competitors
Onchain security exists, but it was designed for humans and wallets. Blockaid (US$ 83M raised, protects MetaMask and Coinbase) and GoPlus Security (about US$ 15M) offer risky APIs for dApps and end users. Manual audits, such as those from CertiK, cover contracts before deployment. Tools such as Flashbots Protect act to send the transaction. None of these products was meant to be called by an AI agent via MCP, in milliseconds, before an x402 payment. Our real competitor today is the agent that executes without any verification.
Competitive differentiation
Blockaid and GoPlus serve humans and DApps, with APIs designed for front-ends. Hedwig was born agent-native: a response in milliseconds via MCP, a price per call that fits an x402 microtransaction, and a verdict designed for automatic decision, not for human reading. Manual auditing takes weeks and covers a contract; we cover any address in real time, at the time of payment.
Entry barrier
Each verification feeds our proprietary base of contracts, scams, and attack patterns labeled based on actual use. The more agents they consult, the higher the detection rate. In addition, integration via MCP becomes part of the customer's payment flow: changing suppliers means rewriting the agent's security pipeline. Accumulated data, combined with deep integration, is difficult to replicate from scratch.
Traction
Functional MVP: MCP server and API running, with contract and liquidity verification in a test environment. No prescription yet. Next milestone: first integrated and paying agents in the x402 ecosystem by the end of 2026.