Productspublished

Scalable Capital Connects ChatGPT, Claude and Grok to Broker Accounts, but Trades Need Approval

The product moves AI from general investment research into a customer’s live account context. Its central control is simple: models can prepare actions, but customers must approve trades and savings plans before execution.

By 3 min read
Scalable Capital Connects ChatGPT, Claude and Grok to Broker Accounts, but Trades Need Approval
Scalable Capital Connects ChatGPT, Claude and Grok to Broker Accounts, but Trades Need Approval

Listen to this story

The audio brief

About 1:33
0:001:33
Read transcript
Scalable Capital has connected ChatGPT, Claude, and Grok directly to customers’ live brokerage accounts—but the models still cannot place a trade on their own. Launched on August twenty-fifth as Agentic Investing, the feature lets an assistant inspect a real portfolio, compare it with a benchmark, calculate a savings plan for a target, and adjust a limit order from a natural-language request. The key boundary is explicit approval: every trade and every savings-plan change must be confirmed by the customer before execution. Deposits and withdrawals remain restricted to Scalable Capital’s app and website. Technically, the connection uses the Model Context Protocol, an open standard for linking AI systems to outside services. Scalable is letting customers choose familiar external models rather than forcing them into one in-house assistant, but it is not granting an autonomous trading mandate. Other brokers are drawing the line differently. Robinhood’s Agentic Trading uses a separate sub-account with its own budget. Interactive Brokers keeps Ask IBKR advisory-only, while Public supports standing instructions for its own agent and separately allows Claude Desktop to connect to a real account through MCP. Scalable says it holds more than sixty billion euros for over one million customers. The question now is whether account-aware AI becomes useful enough for people to bring into repeated investment decisions, even when the final instruction—and the risk—still belongs to them.

Story brief

3 key points

Scalable Capital’s Agentic Investing turns ChatGPT, Claude, and Grok into account-aware investing interfaces without granting them autonomous execution. Launched August 25, the MCP-based feature can analyze holdings, compare benchmarks, calculate savings plans, and modify limit orders, but customers must explicitly approve every trade and savings plan. Deposits and withdrawals stay in Scalable’s app or website. The...

  1. 01

    Scalable says it holds more than €60 billion for over 1 million customers.

  2. 02

    Every trade and savings-plan change requires explicit customer confirmation before execution.

  3. 03

    Deposits and withdrawals remain restricted to Scalable Capital’s app and website.

Scalable Capital has opened a direct path from ChatGPT, Claude and Grok to customers’ live brokerage accounts. The feature, launched Aug. 25 as Agentic Investing, lets those models inspect portfolio information and help prepare investing actions, while keeping the customer responsible for approving trades and savings plans.

That makes the brokerage connection more consequential than a general-purpose chatbot discussing a stock. A customer can ask an assistant to analyze an actual portfolio, compare it with a benchmark, calculate the savings plan needed for a target, or adjust a limit order through a natural-language prompt.

The assistant can act on account context, not execute alone

The integration runs through the Model Context Protocol, an open technical standard for connecting AI systems with external services. Scalable Capital’s design gives customers a choice of outside models rather than requiring the firm’s own assistant, but it stops short of an autonomous trading mandate.

What the customer approval boundary covers

  • Every trade and every savings plan requires the customer’s explicit confirmation before it executes.
  • Deposits and withdrawals remain available only through Scalable Capital’s app and website.
  • The assistant can still analyze holdings and adjust a limit order from a plain-language request.

Brokerages are choosing different control points

Scalable’s confirmation-first approach is one of several emerging designs. Robinhood launched Agentic Trading on May 27, placing an outside agent in a dedicated sub-account with its own budget. The account structure, rather than a per-order confirmation requirement described for Scalable, is its stated constraint on where the agent trades.

Interactive Brokers has kept its Ask IBKR product on the advisory side: it answers questions about a customer’s portfolio, according to the article, without generating trading instructions or connecting to outside models such as ChatGPT or Claude. Public combines two paths: customers can issue standing instructions to its in-house agent to execute trades, move cash and manage risk, and can separately connect Claude Desktop to a real brokerage account through MCP.

A live-account test for a growing AI habit

The launch targets people already using AI frequently. PYMNTS Intelligence found that 10% of consumers use AI for 27 or more tasks per month; the figure was 19% among millennials. The research includes personal investment management among the higher-complexity tasks associated with those heavy users.

For Scalable Capital, which says it holds more than €60 billion for more than 1 million customers, the immediate product question is not whether an assistant can produce a market view. It is whether a model connected to the customer’s real positions is useful enough that people will repeatedly bring it into the decision process, even when the final instruction remains theirs.

Sources

  1. pymnts.comRetail Investing Opens Up to AI Assistants | PYMNTS.com