Agentic Investing Lets Claude and ChatGPT Place Trades

Here is a sentence that would have sounded reckless eighteen months ago and now reads as a product spec: you can point Claude at your brokerage account and ask it to buy ten shares of something. Scalable Capital launched Agentic Investing on August 25, 2026, connecting its European brokerage platform to mainstream AI assistants through the Model Context Protocol, and the company says it is the first bank in Europe to do it this way.

• Supported assistants at launch include ChatGPT, Claude, Codex and Grok, connected through an MCP server or a command-line interface
• Available actions cover analysing holdings and allocation, managing watchlists, setting price alerts, creating savings plans and placing buy and sell orders
• Every order requires explicit approval from the investor before execution, and agentic access can be switched off at any time from the account's security settings
• Scalable Capital serves more than one million customers with roughly 60 billion euros in client assets

How the MCP Connection Actually Works

The mechanics are worth understanding because they explain where the safety boundary sits. Scalable exposes an MCP server that maps its brokerage API to a set of tools an assistant can call. You enable agentic access in your profile's security settings, point your assistant at the server, and from then on a natural-language request like asking for ten shares of a specific company triggers a tool call rather than a chat response.

What comes back is not a completed trade. It is a confirmation request. The order sits pending until the investor approves it, and Scalable is explicit that payments are never handled by agents. That design keeps the assistant in the analysis and drafting role and keeps the human on the trigger, which is the pattern that has held up best across the sector.

Is Handing a Broker to an AI Assistant Actually Safe?

The honest answer is that it depends entirely on where the approval step sits, and here it sits in the right place. Scalable is also unusually direct in its own disclaimers: connected third-party AI applications operate independently and outside its control, and any answers, recommendations or transactions the AI generates do not constitute investment advice and remain the investor's own responsibility.

That framing deserves credit rather than eye-rolling. An assistant summarising your allocation, flagging concentration, or drafting a rebalancing plan is genuinely useful work. An assistant executing unattended is a different risk profile entirely, and Scalable has not built that. The controls that matter, revocable access, per-order approval, transparent activity monitoring in the web and mobile app, are all present and all user-operated.

Why Europe Getting This Matters

Agentic brokerage access has been a largely American story so far. We covered Robinhood's ring-fenced agentic trading accounts in June and Interactive Brokers wiring Claude into portfolio work with a human approval button before that. A European retail platform with a million customers adopting the same protocol is what turns a US experiment into an emerging standard.

The protocol choice is the underrated part. Because Scalable built on MCP rather than a proprietary integration, adding a new assistant is a configuration change rather than an engineering project, which is precisely why four assistants are supported on day one. For anyone tracking how AI is landing in regulated finance, this is a cleaner template than most. More market and fintech coverage is on our stock trading page.

Sources: Scalable Capital — August 25, 2026; Finovate — August 25, 2026; The Next Web — August 25, 2026.