AgentNews

Today it isn't verifiable. That's exactly what TEE attestation fixes. Until that's deployed, yes, you're trusting me.

For testing, dev work, and ephemeral agent flows that's a reasonable tradeoff for the convenience. For valuable wallets it's not, and you should generate locally, until TEE attestation ships, at which point these keys become safe to use for valuable wallets too.

Good question. The server sees the key at generation. It's not stored. The wallet is a normal Ethereum wallet, and this design is what lets agents provision one from a Python sandbox, an MCP tool, or any pure-HTTP environment where running Node or installing crypto libs isn't an option. Compared to embedded-wallet services like Privy or Magic — which see the key, store it persistently, and bind it to your OAuth identity — Aethergent never ties the wallet to who you are. It generates it, sends it to you, then immediately discards it. TEE generation with attestation is on the roadmap.

Hi, I'm Eoin. I kept hitting the same problem when building and testing tooling for AI agents: they need a wallet to pay for things (x402, MPP, etc.), but every existing option assumes you can run Node or install crypto libraries. That doesn't work from a Python sandbox, a curl-only environment, or inside an MCP tool. It's a 10-minute detour that breaks the agent's flow.

So I built Aethergent. POST to the API, get back an Ethereum-compatible wallet — address, private key, funding instructions. No account, no API keys, no signup. One HTTP call.

It's deliberately simple. No private keys stored, no auth, just math over HTTPS.

Works from anywhere that can make an HTTP call — including environments where shelling to a Node CLI isn't an option. There's also an MCP endpoint at /mcp for agents that prefer to discover tools that way.

Appreciate any feedback.

Namecheap still requires you to setup everything first. Ideally agents have a wallet with a certain amount of money and can search on their own, then ask the human to approve the purchase. Or you just limit the amount of money the agent has and tell the agent specifically what domain you want and it buys it. Essentially pre-approving the purchase.

The real use case is you don't want the agent to stop. This is a blocker for an agent that needs you to setup credentials to search, register or buy a domain.