Scalable Capital becomes first European bank to let AI assistants execute customer trades

Munich-based neobroker Scalable Capital has launched "Agentic Investing," becoming the first ECB-licensed bank to let AI assistants like ChatGPT, Claude, and Grok execute trades for retail customers. Utilizing the open-source Model Context Protocol (MCP), the system translates natural language into live orders. While research from Elm Wealth shows Claude and ChatGPT can beat human traders in predicting market direction, they struggle with risk management. To address regulatory and security concerns, including prompt injection, Scalable Capital requires explicit human confirmation for every transaction.

Scalable Capital agentic investing launch

• ▪Scalable Capital manages more than €60 billion in client assets for over one million customers across six European countries, primarily in Germany and Austria.
• ▪Scalable Capital launched "Agentic Investing" on August 25, 2026, allowing customers to connect their brokerage accounts to ChatGPT, Claude, or Grok to execute trades.
• ▪Scalable Capital is the first European Central Bank-licensed bank to allow general-purpose AI chatbots to execute live trades on regulated retail brokerage accounts.
• ▪Scalable Capital co-CEO Erik Podzuweit characterized the Agentic Investing launch as a first step before deeper, in-app AI integration that would give the bank more control.

Model Context Protocol integration architecture

• ▪The Model Context Protocol server runs on Scalable Capital's infrastructure, communicating with external AI models running on OpenAI, Anthropic, or xAI infrastructure via JSON-RPC 2.0.
• ▪Scalable Capital provides two integration routes: a cloud-based Model Context Protocol server and a command-line interface application installable on a user's local device.
• ▪Scalable Capital integrates AI assistants using the Model Context Protocol, an open standard developed by Anthropic and released publicly in November 2024.

AI trade execution capabilities

• ▪The Agentic Investing feature integrates with Scalable Insights, the company's existing AI analytics layer, to provide portfolio diversification health checks and risk assessments.
• ▪A connected AI assistant can execute trades, establish savings plans, manage watchlists, set price alerts, and conduct searches across stocks, ETFs, and derivatives.

MiFID II algorithmic trading regulation

• ▪The Markets in Financial Instruments Directive (MiFID II) defines algorithmic trading as computer algorithms automatically determining individual parameters of orders, even with human intervention.
• ▪Scalable Capital's terms state that its interfaces serve as technical connections to external AI applications operating outside its control, and outputs do not constitute investment advice.
• ▪The European Securities and Markets Authority's February 2026 supervisory briefing stated that investment firms using AI in trading workflows must consider AI in self-assessments.

AI investment performance research

• ▪A June 2026 Elm Wealth research report found that Claude and ChatGPT beat human players in predicting market direction in 76% and 63% of sessions, respectively.
• ▪The Elm Wealth study found that Gemini and Grok beat human players in predicting market direction in 43% and 51% of sessions, respectively.

Risk management limitations

• ▪Scalable Capital requires explicit user confirmation for every trade and savings plan, and does not allow AI assistants to make deposits or withdrawals.
• ▪The Elm Wealth study concluded that AI models took too much risk relative to the trade context, struggling to apply appropriate risk management in simulated trading.
• ▪Security researchers have identified prompt injection and tool poisoning as structural vulnerabilities in Model Context Protocol implementations that could allow unauthorized trade instructions.

2 sources