CoinFello Debuts Publicly at EthCC 2026, Bringing AI-Powered DeFi to Retail Users

• CoinFello has launched publicly at EthCC 2026, opening its AI-powered DeFi interface to users beyond its private alpha.
• The platform combines natural-language onchain execution with self-custody, delegated permissions and smart account automation.

CoinFello has officially gone public, using EthCC 2026 in Cannes as the stage for a launch that speaks to a familiar problem in crypto. Plenty of people hold digital assets. Far fewer actually use DeFi.

The company, based in Fort Worth, Texas, says its product is built for that gap. After a period in private alpha, CoinFello is now available through a web app and a newly released Android app, with iOS expected later. Its pitch is straightforward enough. Let users send tokens, swap, bridge, and automate on-chain actions through a chat interface, without giving up custody.

A conversational layer for wallets and DeFi

CoinFello describes itself as a self-sovereign AI agent for on-chain activity. In practice, that means users can type plain-language prompts instead of jumping between dApps, bridges and wallet pop-ups. The platform then translates those instructions into blockchain transactions and presents them in readable terms before anything is signed.

That is clearly aimed at the crowd sitting on idle tokens, not because they lack interest, but because onchain execution still feels like a minefield. One wrong click, one bad route, one approval too broad, and the cost can be real.

Delegated permissions, not blind wallet access

Security sits at the center of the launch message. CoinFello says it does not hand AI agents full wallet control. Instead, it uses delegated permissions with configurable limits on token amounts and timeframes. Private keys stay on the user’s device, with hardware-level protection on macOS through Secure Enclave.

The platform also plugs into the wider AI agent stack. Through integrations with tools such as OpenClaw, Claude Code, Kiro and Windsurf, CoinFello can function as an execution layer for personal agents that need to perform onchain actions within user-set boundaries.