CoinFello has gone public with a pitch that’s becoming increasingly common in crypto: make DeFi usable for people who already hold assets but never interact with it. The platform, unveiled at EthCC 2026 in Cannes, introduces a conversational AI agent that lets users execute onchain actions through simple prompts — without giving up custody of their wallets.

It’s a direct attempt to solve a long-standing gap in the market. While hundreds of billions of dollars sit in DeFi protocols, participation remains relatively narrow. A large portion of crypto holders still avoid staking, bridging or yield strategies, not because of lack of interest, but because of complexity and risk.

CoinFello’s bet is that lowering that barrier — without introducing new security trade-offs — could unlock a much larger user base.

What CoinFello actually does

The platform offers a chat-based interface that translates natural language into blockchain transactions. Users can send tokens, swap assets, bridge across networks or interact with DeFi protocols by typing commands instead of navigating multiple interfaces.

Unlike earlier wallet automation tools, CoinFello is trying to bundle research, execution and automation into one flow. That means a user could ask for a yield opportunity, evaluate it and execute the transaction within the same interface.

The product is now live via a web app and Android application, with iOS expected later.

Investor Takeaway

The real opportunity is not new users — it’s inactive capital. Platforms that can activate idle crypto holdings could drive the next wave of DeFi growth.

How CoinFello handles security

Security remains the main challenge for AI-driven trading and automation tools. Giving an agent full wallet access is an obvious risk, and most users are not willing to hand over private keys.

CoinFello’s approach is based on delegation. Instead of direct control, users assign limited permissions to the system — defining how much can be spent, for how long and under what conditions.

Private keys stay on the user’s device, and on supported systems are protected at the hardware level. Transactions are also presented in human-readable form before execution, giving users a final approval step.

This model sits somewhere between manual control and full automation, which may be necessary for wider adoption.

Why AI agents are moving into DeFi

The timing of the launch reflects a broader shift. AI agents are starting to move from simple assistants into tools that can take action — monitoring markets, executing trades and managing positions.

In DeFi, that trend is particularly relevant. The space is fragmented, technical and often unforgiving. A single mistake can lead to permanent loss, which keeps many users on the sidelines.

By introducing a layer that abstracts complexity while preserving control, projects like CoinFello are trying to make DeFi behave more like traditional financial apps — but without removing decentralization entirely.

The company is also positioning itself as infrastructure for other AI systems. Through integrations with agent frameworks, users can connect personal AI agents and allow them to execute transactions within predefined limits.

Investor Takeaway

Execution layers for AI agents could become a key part of the DeFi stack. The question is which platforms users trust to actually move funds.

What comes next

The public launch marks the first phase of CoinFello’s rollout, with a broader roadmap focused on deeper automation and more decentralized infrastructure.

Short term, adoption will depend on usability and trust. Users need to feel confident that the system can execute transactions correctly without introducing new risks. Longer term, the opportunity is larger: turning passive holders into active participants.

If even a fraction of dormant crypto capital starts moving through DeFi, the impact on liquidity and product usage could be significant.

For now, CoinFello is entering a crowded but fast-moving space. AI-driven interfaces are becoming a common narrative. The difference will come down to execution — and whether users are willing to let software act on their behalf, even with guardrails in place.