Funds, token swaps, balance checks, and transaction history reviews will be enabled for AI models through a new tool launched by Base.
A tool has been launched by Base, the Ethereum layer-2 blockchain developed by crypto exchange Coinbase, to connect Base accounts with artificial intelligence agents for blockchain operations.
Base said on Tuesday that its new Base MCP (Model Context Protocol) allows users to instruct AI agents such as Claude or ChatGPT to transfer funds, swap tokens, check balances, review transaction history, and use supported apps within the ecosystem.
The tool allows users to manage their crypto directly through an AI model’s chat interface and is also able to interact with crypto protocols such as Morpho, Moonwell, Uniswap, Aerodrome, Avantis, Bankr, and Virtuals, Base said.
An action is first proposed by the AI agent in the chat, after which the Base wallet is opened in a new window where users can confirm or cancel the transaction. The agent is not granted access to private keys, and every action must be confirmed by the user.
Agentic payments have been touted as the next major use case for crypto, as backers argue that the banking system may be difficult for AI models to access and that digital assets will be required for transactions.
It was stated by Base that every transaction proposed by the agent passes through the same review flow used for any Base account request, with asset changes being simulated before a user confirms.
Lincoln Murr told Fortune that, unlike siloed agentic wallets that only exist in a terminal, a Base Account “travels with you — trades, history, and portfolio sync whether you’re in-agent or in the Base App,” as stated.
Adoption of the Base MCP is expected to be expanded alongside the Coinbase x402, an agentic AI payment standard launched by the company in May 2025.
Lincoln Murr described MCP as a “nice wrapper” built on top of APIs. Alongside Coinbase x402, it enables a new micro-transaction economy in which agents are able to make small crypto payments.
However, that economy remains in its early stages, with only $1.1 million in volume processed by Coinbase x402 over the past 30 days, according to x402scan.
The use of AI agents for crypto payments has also drawn criticism, with a recent research paper from Google and leading universities arguing that AI agents should be treated as an untrusted system component.
The researchers said AI agents should clearly separate instructions from untrusted data to prevent attackers from deceiving the agent with hidden malicious commands.
Earlier this week, Socket discovered malware targeting crypto developers by injecting hidden instructions designed to hijack AI coding assistants.


