Questrade Connects Brokerage Accounts To AI Agents

What Questrade MCP Can Do

A general chatbot can discuss markets. Questrade MCP can work with the investor's authorized account data. Once connected, the agent can retrieve approved balances, positions, orders and market information, then combine that context with outside research.

Questrade's live documentation describes both read and write access, although "write" doesn't mean autonomous trading. The agent prepares the instruction. The customer sees the symbol, buy or sell direction, order type, quantity, price, estimated cost and commission, then approves it through Questrade. The agent can't skip that review or place, change or cancel the order on its own.

Customers choose the accounts and permissions. They can share data without allowing trade drafting, and they can revoke the connection through Questrade's API settings. There is an important catch.

Questrade says revocation stops future access but does not remove information already shared with the AI provider. Closing the connection and retrieving previously shared data are two different things.

The connection becomes more interesting beside Custom Indexing. Launched in June, the product uses templates, fractional shares and one click rebalancing across as many as 600 securities. Questrade now says an agent can help build a custom index too. A conversation can therefore produce a structured group of proposed orders, with the customer still deciding whether to proceed.

Questrade MCP Beta Arrives As CIRO Sets AI Expectations

The timing is hard to ignore. In March, CIRO guidance gave execution only dealers more room to provide tailored education, alerts and other decision support. The core restriction remains. These dealers cannot recommend a security.

Questrade MCP puts that boundary in front of a real customer. The agent can discuss actual holdings and turn a plain language request into a draft order. Questrade says the analysis, orders and custom indices come from the agent and are not its advice or recommendations. The customer is expected to check the work and approve the result.

The disclaimer doesn't settle every responsibility. Questrade still controls authentication, account permissions, the order screen and execution. CIRO's 2026 compliance report says examiners will ask how dealers use AI and test the controls around it. The IOSCO AI toolkit raises similar questions about accountability, data, monitoring and third party providers. Neither regulator has publicly approved Questrade MCP as a product.

Operators will want to see the working details. When does a prompt become an instruction? When might an answer sound like a recommendation? What happens if the agent uses stale information, mistakes a ticker or drafts the wrong order type? A clean record of the request, response, draft and final approval will be just as important as the chat experience.

How Questrade Compares With AI Trading Platforms

Questrade enters an active race. Interactive Brokers connected accounts to Claude, ChatGPT and Grok in June. Its customers can research markets and prepare instructions for equities, ETFs, options and futures, but they approve every trade. That is the closest public comparison to Questrade's current model.

Robinhood has gone further in the United States. Its Agentic Account gives a third party agent a dedicated budget and allows trades without direct approval each time. Robinhood warns that the agent can make mistakes, customers can lose their entire investment and the customer remains responsible for monitoring the account.

Questrade's version is closer to assisted execution. The agent can see the authorized account and prepare the action, while Questrade keeps the approval gate. It gives up some autonomy in exchange for a visible decision before each order.

The contest isn't limited to AI connectors. Wealthsimple and bank owned brokerages are already competing to become the main Canadian investment account. Questrade is assembling zero commission trading, real time fractional shares, Custom Indexing and agent access while its parent prepares Questbank. We don't yet know whether customers will use these products together or treat them as separate features.

What Makes Questrade Different In Summer 2026

Where it could work A customer can ask about a portfolio, research an idea and prepare an order without copying balances and tickers between several screens. Custom Indexing could make that especially useful for people managing many positions.

Where it could get messy The account data may be current while the outside information is incomplete. An agent can misunderstand an instruction or produce analysis that a customer hears as advice. The approval screen helps before execution, but it can't resolve every question involving data retention, explanation or customer understanding.

Questrade already has regulated execution, a large Canadian customer base and more than C$80 billion in assets under administration. The open question is usage. If customers repeatedly research and prepare orders through an agent, MCP becomes another front door to the brokerage. If they mostly ask for balances, it remains a convenient connector.

Talking Point

When an AI agent can read a portfolio and prepare a trade, which controls keep useful assistance from becoming an unintended order or advice the dealer isn't allowed to give?