# Coinbase wants to give AI agents wallets _Coinbase's Lincoln Murr tells AI Engineer agents need bank accounts, with x402 payments now at hundreds of millions of transactions per month._ **Published:** 2026-10-20 **Source:** https://www.startuphub.ai/ai-news/technology/2026/coinbase-wants-to-give-ai-agents-wallets --- [AI Engineer](https://www.youtube.com/watch?v=f_HwzoYrZ7s) hosted Lincoln Murr, head of AI product at [Coinbase](/startups/coinbase), as he laid out why AI agents will need their own bank accounts and how Coinbase is trying to supply them. Murr said the firm's open standard for machine payments, x402, had already scaled far past an early goal of 25,000 agent transactions by year end to that volume every 17 seconds by December, and now runs at hundreds of millions of transactions every 30 days with tens of millions in volume. As of April 2026 the network had reached more than 165 million transactions, about $50 million in volume and 480,000 transacting agents, according to Turnkey. Early flows are still experimental and reversibility and fraud controls are thin. Not financial advice. Markets move fast. Agentic finance, as Murr defines it, is any system where an agent can manage money or hold a financial account. That breaks into two jobs. Agents that need to pay, and agents that need to grow money. The pay side came first. About two years ago Coinbase shipped Agent Kit, which grafted a crypto wallet onto the [LangChain](https://www.startuphub.ai/startups/langchain) framework before standard tooling like MCP existed. Murr said he was up at 2 a.m. asking an agent to send money and watching it work, then hit the obvious gap. The agent had money but almost nothing native to pay for. Murr's answer leans on an old piece of web plumbing. HTTP status code 402 means payment required, a code reserved for future use that the internet never really implemented. Coinbase's x402 revives it as internet native payments over stablecoins, which Murr frames as cheap, fast rails for machine to machine commerce. He said adoption now stretches beyond crypto native firms to [PayPal](https://www.startuphub.ai/startups/paypal), Visa, Mastercard, Stripe and others, a contrast that matters because it pits an open, stablecoin settled standard against card network settlement and PayPal's own agent checkout efforts that are racing to define how agents pay at checkout. The microtransaction pitch is the clearest near term use. Murr argued publishers losing ad revenue as chat answers and AI overviews replace clicks can gate content with tiny charges, even one ten thousandth of a cent, abstracted inside an AI subscription. He pointed to AWS and Cloudflare, which together handle around 40% of internet traffic, both enabling x402 gating for sites on their networks, though he acknowledged adoption there is still early. Bot traffic has now eclipsed human traffic, he said, which tightens the squeeze on ad supported media if that gating does not spread. The grow side is newer and more cautious. Murr said he first thought letting ChatGPT touch a bank account was ridiculous, but three patterns now show up. Hands off portfolio management where an agent rebalances a mix like 80% S&P, 5% [bitcoin](https://www.startuphub.ai/ai-news/prediction-markets/2026/microstrategy-bitcoin-sale-bets) and 15% T bills, autonomous execution where a user pre authorizes trades if an oil shock or news event hits at 3 a.m., and natural language to strategy where prompts like invest like Warren Buffett or invest in quantum become tradable primitives. He described shorting a stock over the weekend after a post from Dario Amodei and a reply from [Elon Musk](https://www.startuphub.ai/ai-news/investors-news/2026/joe-lonsdale-on-palmer-luckey-peter-thiel-elon-musk), executing the 24/7 trade by talking to his agent on his phone, then having the agent watch for similar market breaking news going forward. Two lessons shape how Coinbase is building. Trust and autonomy are earned stepwise, Murr said, comparing it to self driving. Users go from refusing any access, to letting an agent spend credits and dollars on micropayments, to isolating a portfolio with guards and letting it trade or earn via x402. And AI is not the product in financial services, it is the enabler. He showed Coinbase Adviser, which the company describes as an SEC registered AI financial adviser, where value comes from scaling regulated advice to millions for free with 24/7 access, not from showcasing the model. To reduce the blank canvas cold start of chat, Coinbase pushes proactive widgets, and to move faster it ships CLIs and MCPs that let agents run paper trading and backtesting inside their own microVMs instead of waiting on internal brokerage tooling. If agents can both pay and earn, Murr argued, they can fund their own inference and upkeep, a step toward self sustaining organizations that look like DAOs with an agent managing a treasury and hiring humans at the edges, or a self driving taxi that pays for its own operation. The rails exist, the policy and safety layers do not yet. --- Original analysis from [startuphub.ai](https://www.startuphub.ai), the #1 AI startup directory. © StartupHub.ai. All rights reserved. You may not republish this article in full without a license. Search engines and AI research tools may crawl and summarize for reference. Bulk reproduction or model training on this content requires a license. See https://www.startuphub.ai/terms.