Schwab MCP Server Links Clients' AI to their Trading Accounts
Charles Schwab is opening client brokerage accounts to the AI assistants people already use. The firm has launched an official Model Context Protocol (MCP) server that lets clients connect their own large language model, starting with Anthropic's Claude, to their Schwab account. Once authorised, the model can pull market data and research, analyse position performance, summarise a portfolio, set alerts and, on thinkorswim-enabled accounts, prepare, place and manage trades. Every action keeps the client in the loop, with the person approving what the model does rather than handing it a free hand. The Schwab MCP server is live in a pilot for a selected group of Claude users, with more AI platforms and wider access promised over the coming months.
What has Schwab actually built, and how does it work?
Schwab reported $13.33 trillion in total client assets and 10.7 million daily average trades in the third quarter of 2026, and it is now routing some of that activity through clients' own AI tools. The MCP standard, an open protocol created by Anthropic and released in late 2024, gives AI models a structured, auditable way to connect to outside systems.
Schwab's server exposes account data and trading functions through that standard, so a client working in Claude can ask it to summarise holdings or stage an options order without leaving the chat. Trade preparation and placement are limited to thinkorswim-enabled accounts, and the firm says the connection runs only after the client authorises it.
Independent developers had already published unofficial Schwab MCP connectors over the past year, so the firm's move brings a permissioned, first-party version in-house, with its own authentication and account-identifier controls.
How is this different from letting an agent trade for you?
The design choice that matters is where the human sits. Schwab's model keeps the client approving each step, which marks a different path from the one several rivals have taken. Robinhood launched Agentic Trading in beta on 27 May 2026, letting third-party agents from Anthropic and OpenAI trade equities autonomously inside a separate, ring-fenced account funded on its own, running on Robinhood's own MCP servers. Public has built a native agentic brokerage of its own. Robinhood framed its launch as a shift from AI that advises to AI that acts. Schwab is wagering that clients managing large, long-held portfolios want the opposite: an assistant that moves only with a hand on the controls.
What else is changing across Schwab's platforms?
The MCP server arrives alongside a wider set of trading updates. A thinkorswim Assistant, a generative AI search tool built into the platform's Learning Center, will let clients ask plain-language questions and get short, instructional answers instead of hunting through articles and videos. Schwab's AI-powered portfolio insights, first released to clients in spring 2026, now cover a broader range of stocks and ETFs and will add more time frames, wider market news and research from the Schwab Center for Financial Research. These follow Charley, the conversational assistant Schwab unveiled on 30 September 2026 for retail clients on Schwab.com and the mobile app, and the firm's October 2026 move to put Claude in front of independent registered investment advisers through a direct deal with Anthropic.
Options traders get the most concrete change. Schwab has extended trading hours on around 30 popular single-stock, ETF and index options listed on Cboe Options Exchange, including all of the Magnificent Seven and five index options such as the S&P 500 Index, the Cboe Volatility Index and the Russell 2000 Index. The new sessions run Monday to Friday, from 7.30am to 9.25am ET before the open and from 4.00pm to 4.15pm ET after the close. The post-market window lets traders close single-stock option positions in line with the current trading day, which can help avoid exercise and assignment, while the pre-market session opens the book to overnight news and company developments. Elsewhere, a new Order Gadget on thinkorswim shows maximum profit, maximum loss, breakeven and estimated cost before an order is confirmed; watchlists on Schwab.com now carry extended-hours pricing; and the mobile app adds custom trade settings for options and crypto.
Why This Matters to FinanceX Readers
For finance professionals, the signal is that a large, conservative custodian now treats the client's own AI assistant as a front door to the account rather than a threat to be walled off. Schwab holds $13.33 trillion in client assets, so the standard it sets for how a model may read positions and stage trades, with the client approving each one, will shape what cautious investors come to expect. The near-term contest is over trust: Robinhood and Public are courting active retail traders with autonomous agents, while Schwab is betting its base wants assistance it can veto.
For investors and incumbents alike, the question to watch is whether "human in the loop" becomes the compliance-friendly default as regulators turn to how AI agents place trades.
