The development team behind the artificial intelligence agent platform CoinFello has introduced an open-source tool designed to allow AI agents to securely interact with crypto wallets and perform blockchain transactions. The newly released skill enables AI systems to connect directly to the widely used wallet service MetaMask while ensuring that users maintain full control of their private keys.
According to the announcement, the new capability allows personal AI agents built on the OpenClaw framework, commonly referred to as MoltBots, to initiate transactions using funds stored in a user’s existing MetaMask wallet. The system permits agents to operate with a predefined allocation of cryptocurrency without requiring wallet owners to relinquish control over their private keys or overall account custody.
The company indicated that the technology was designed to address one of the primary security concerns associated with AI-powered blockchain tools. Many existing AI wallets rely on direct access to private keys or API credentials, creating potential vulnerabilities that could be exploited through prompt injection attacks or other manipulation techniques.
Smart Account Infrastructure Enhances Security
CoinFello’s system was built using the MetaMask Smart Accounts Kit, which integrates blockchain account abstraction technology. The solution utilizes the Ethereum standards ERC-4337 and ERC-7710 to enable secure transaction delegation.
Through this architecture, users can grant narrowly defined permissions to AI agents. Rather than providing full wallet access, individuals can authorize specific tasks, transaction limits, or actions that an AI agent is allowed to perform. This permission-based structure significantly reduces the risk of unauthorized activity.
The company explained that the model was designed to ensure that AI agents operate only within clearly defined boundaries. Instead of having unrestricted control over a wallet, the agents function with temporary or task-specific permissions that limit their operational scope.
CoinFello’s founder and chief executive officer, known publicly as Jacob C, previously served as the head of operations for MetaMask. His experience within the wallet ecosystem influenced the design of the new tool, which focuses on strengthening security while enabling more advanced automation within blockchain systems.
Building a Safer AI-Driven On-Chain Economy
The company’s leadership emphasized that improved security frameworks are essential if AI agents are expected to play a meaningful role in the decentralized economy. The organization explained that traditional models, in which autonomous systems are given full access to private keys, present unacceptable risks for users and blockchain infrastructure.
CoinFello’s chief technology officer indicated that the company believes meaningful AI participation in the on-chain economy requires security models that go beyond simply handing an autonomous system a wallet’s private key. The CTO explained that the platform’s approach attempts to create a safer interaction model by limiting the permissions granted to AI agents and ensuring that wallet ownership remains fully with the user.
Although MetaMask did not release a formal statement following the announcement, earlier signals from its development team suggested support for the concept. A product lead from Consensys responsible for the MetaMask Smart Accounts Kit had previously expressed satisfaction about collaborating with CoinFello to enable agent-driven blockchain experiences. The representative indicated that the long-term vision involves AI agents performing blockchain actions safely on behalf of users while relying on granular permission structures that allow individuals to control how transactions are executed.
Wide Range of Blockchain Capabilities
The AI agent skill supports a broad set of blockchain activities. Among the functions available to MoltBots are token swaps involving ERC-20 assets, cross-chain bridging between Ethereum Virtual Machine networks, non-fungible token interactions, staking operations, lending protocols, and complex multi-step trading strategies.
— CoinFello (@CoinFello) March 12, 2026
These activities can be triggered through natural-language prompts, allowing users to instruct their AI agents using conversational commands rather than manual transaction management.
CoinFello confirmed that the tool has been released under the MIT open-source license, enabling developers to freely build upon the technology and integrate it into their own blockchain-based AI systems.
Growing Momentum for AI Agent Ecosystems
The launch comes during a period of rapid growth for the OpenClaw ecosystem and its associated MoltBot agents. Interest in AI-powered blockchain activity has increased significantly in recent months, driven partly by the popularity of a social platform called Moltbook.
The platform, which is primarily populated by autonomous OpenClaw agents, experienced viral adoption earlier this year and contributed to record levels of token activity on the Base blockchain launchpad known as Clanker.
Recent industry developments have also highlighted the rising importance of AI-driven platforms. Media reports indicated that Meta Platforms, the parent company behind major social networks such as Facebook, Instagram, and WhatsApp, had acquired Moltbook and brought its founders into the company’s artificial intelligence division.
Against this backdrop, CoinFello’s release demonstrates how developers are working to combine AI automation with blockchain infrastructure while addressing the critical security challenges associated with autonomous financial systems.