ThinkMarkets Launches MCP Server: AI Can Execute Trades, but Not Access Funds

ThinkMarkets, the global CFD and forex broker headquartered in Melbourne and London, has launched a Model Context Protocol (MCP) server named ChelseaAI, the company's co-founder and CEO Nauman Anees revealed to Finance Magnates. The server allows traders to connect any AI large language model (LLM) of their choice directly to ThinkMarkets' proprietary ThinkTrader platform, enabling order execution without the need to log into the trading interface.

'This means you no longer need to worry about charting, automated trading, indicators, or analysing market data. Your AI handles all of it,' Anees said, noting that the move removes significant user friction. However, he stressed that the AI cannot access traders' funds or make deposits or withdrawals – a limitation designed intentionally for security. ThinkMarkets has implemented a permissions system called 'scopes' that lets traders control whether the AI can execute orders, and set limits on order types. Circuit breakers and hard limits prevent the AI from executing the entire account on a single trade or margin.

ChelseaAI is exclusive to ThinkMarkets' proprietary ThinkTrader platform and is not supported on third-party trading platforms like MetaTrader. 'We can control a lot of the permissions and scopes on our proprietary platforms. That's not something we can do, obviously, on other external platforms,' Anees explained. The broker holds licenses in multiple jurisdictions including Australia, the UK, Cyprus, South Africa, Dubai, New Zealand, and Japan, and the MCP server will be available globally to all clients.

ThinkMarkets joins a growing trend among brokerages embracing MCP servers. Spotware, the developer of cTrader, recently launched a similar server, while eToro offers 'Agent Portfolios' for linking AI agents. IG Group's Australian division launched a read-only MCP server for ChatGPT, and Robinhood has enabled AI agents to trade stocks and make payments. Anees expects more industry players to attempt similar launches but cautioned that deep integration – including data security, circuit breakers, and agentic controls – requires significant proprietary technology.

Data privacy remains a key concern, particularly under GDPR in Europe. Anees emphasised that user data never leaves the account ecosystem; each session is closed without tracking, and ThinkMarkets does not move data. 'If we move the data, then it's a separate topic, and we never do that,' he said. The CEO also noted that the value of an interface is diminishing as APIs, MCPs, and cloud infrastructure become more critical, allowing traders to focus on strategy rather than manual chart analysis.