Base’s Model Context Protocol server just got a lot more useful for the machines. The team added 13 new app skills to Base MCP, the open-source toolkit that lets AI agents interact with onchain applications through a standardized interface. The new capabilities span trading, lending, NFT creation, token launches, and—critically—x402 payments, the HTTP-based protocol that lets agents pay for off-chain services with on-chain stablecoins. This is not a theoretical roadmap item. The skills are live, documented, and composable, meaning a developer can now wire an agent to swap USDC for ETH on a DEX, supply it to a lending pool, mint an NFT with the yield, and pay a data provider for the price feed that triggered the whole sequence—all without a human in the loop.
What Base MCP Actually Does
MCP, or Model Context Protocol, is an open standard originally developed by Anthropic that gives large language models a structured way to call external tools. Base’s implementation wraps onchain actions—smart contract interactions, transaction construction, wallet operations—into MCP tools that an AI agent can invoke through natural language or programmatic logic. Before this update, the toolkit was limited to basic reads and a handful of write operations. The 13 new skills change that. They include token swaps via decentralized exchanges, supplying and borrowing on lending protocols, minting NFTs, deploying meme tokens, and interacting with the x402 payment standard. Each skill is a self-contained module with defined inputs, outputs, and error handling, so an agent can reason about which tool to use and in what order. The architecture matters because it abstracts away the chain’s complexity. The agent does not need to know about gas limits, nonce management, or ABI encoding. It just says “swap 100 USDC for ETH” and the MCP server handles the rest.
The x402 Connection: Payments Without Wallets
The most consequential addition for the stablecoin economy is x402 support. x402 is a payment protocol that uses the HTTP 402 status code—“Payment Required”—to let an API or web service demand onchain payment before delivering content. An AI agent hitting a paywalled endpoint receives a 402 response with a payment address and amount. The agent’s MCP tooling constructs the transaction, broadcasts it, and retries the request with a proof of payment. With Base MCP now natively supporting x402, an agent can autonomously pay for data feeds, API calls, inference compute, or any digital good priced in USDC on Base. This closes a loop that has been mostly theoretical until now. We have written before about x402 as a primitive for agent-to-service payments, but the missing piece was the agent-side tooling to actually execute those payments without custom integration work. Base MCP provides that tooling as a standardized, reusable component.
Composability Is the Real Story
Individually, each of the 13 skills is a useful primitive. But the design philosophy behind MCP is that tools should be chainable. An agent can use the output of one skill as the input to another. A simple example: an agent monitoring a lending protocol detects that borrowing rates for USDC have spiked. It uses the swap skill to convert idle ETH to USDC, the lending skill to supply that USDC to the pool, and the x402 skill to pay a notification service that alerts a human operator. More complex workflows are limited only by the agent’s reasoning capacity and the developer’s imagination. This composability is what distinguishes Base MCP from single-purpose bot frameworks. It is not a trading bot or a payment processor. It is a general-purpose onchain action layer for AI agents, and the 13 new skills make it general enough to cover most of the economic activities that matter on a smart-contract chain.
What This Means for Agentic Commerce
The expansion of Base MCP signals a shift in how Coinbase and the Base ecosystem think about AI agents. Early experiments focused on giving agents wallets and letting them hold or transfer assets. That is table stakes. The new skills turn agents from passive asset holders into active economic participants that can trade, lend, create, and pay. For stablecoin issuers and payment companies, the implication is that the demand side for onchain dollars is about to get more automated and more programmatic. An agent that can autonomously manage a treasury, pay for services, and deploy capital into yield-generating protocols is a net new category of stablecoin user—one that operates at machine speed and scale. The infrastructure is still early, and the agent frameworks are still finding product-market fit, but the direction is clear. Base is building the rails for an economy where the customer is code.