MCP Protocol Drives Brokers' AI Transformation
On May 28, 2026, IG Group's Australian division announced that it had listed itsCFDassistant on the ChatGPT app store, enabling read-only access to trading accounts based on theMCP. This move came just over ten days after Spotware, the developer of cTrader, released its official MCP server, and less than a year after eToro launched Agent Portfolios. The rapid entry of multiple brokers into this field indicates that the financial trading industry's adoption of the MCP protocol is moving from technical experimentation to large-scale implementation. (Source: Finance Magnates, May 28, 2026)
The MCP protocol was officially released by Anthropic in November 2024 (seeIntroducing the Model Context Protocol). Its core purpose is to establish standardized, secure, two-way connections between AI models and external data sources and tools. The protocol underwent several important updates in 2025, including support for asynchronous operations, stateless improvements and server-side authentication mechanisms. The latest specification version is dated June 18, 2025. (Source: Anthropic official, November 2024)
The Technical Leap from Read-Only Queries to Agentic Trading
Among brokers and trading platforms that have integrated MCP, technical approaches have diverged significantly. IG Australia has chosen the most cautious read-only model, under which the AI assistant can only query data and cannot execute any trading instructions. By contrast, eToro's Agent Portfolios function allows AI agents to execute actual trades within user-defined sub-portfolios, with a minimum threshold of USD 200. (Source: eToro official) cTrader provides differentiated permissions through both remote and local MCP server modes, with the local mode allowing deeper platform control capabilities. (Source: FXStreet, May 18, 2026)
This divergence essentially reflects different judgments about the safety boundaries of AI trading:
Read-only model (IG Australia): zero trade execution risk, with functionality limited to information queries and data analysis
Restricted agentic trading (eToro): AI operates within a defined sub-portfolio, with limits on capital and strategy parameters
Deep integration model (cTrader, Bybit, Gemini): provides a full API layer, allowing AI to execute, manage and analyze trades
The cryptocurrency sector has moved more aggressively. Bybit has released an official MCP protocol, allowing traders to access real-time market data, execute trades and manage portfolios through AI assistants such as ChatGPT and Claude, without writing custom API code. Gemini exchange's Agentic Trading function has also connected its full trading API with the MCP protocol. (Source: BlockBeats)
Comparison of MCP Integration Models Among Major Trading Platforms
| Platform | MCP Integration Model | AI Trade Execution Permission | User Access Requirement |
|---|---|---|---|
| IG Australia | Read-only data queries | No trading operations supported | Hold an IG Australia trading account |
| eToro | Agent Portfolios | Supported (execution within defined sub-portfolios) | Minimum deposit of USD 200 |
| cTrader (Spotware) | Dual remote and local MCP servers | Supported (local mode provides deeper control) | Install the cTrader trading platform |
| Bybit | Official MCP protocol integration | Supported (covering full trading functionality) | Register a Bybit account |
IG Group's Financial Performance and AI Strategy Alignment
IG Group's MCP deployment is not an isolated technical experiment, but aligns with its broader business growth trajectory. According to publicly disclosed data, IG Group's organic total revenue for the first quarter of fiscal 2026 increased 19% year on year to GBP 331.2 million, while reported total revenue rose 21% year on year to GBP 339.9 million. The core growth drivers came from active performance in stock trading and cryptocurrency businesses. (Source: TradingView/Investing.com, fiscal 2026 first-quarter earnings data)
Revenue Growth Supports Investment in AI Infrastructure
Driven by quarterly earnings that exceeded expectations, IG Group's share price rose by approximately 9% to 11% after the results were released, reaching a record high. The decision to increase investment in AI infrastructure while the share price was at elevated levels reflects management's confidence in technology-driven long-term growth. IG Australia has taken the lead in testing an MCP server, using a read-only model to reduce compliance risk while leaving room for a gradual expansion of AI trading permissions in the future.
From an industry signaling perspective, eToro CEO Yoni Assia made remarks in a media interview in May 2026 (Source: Yahoo Finance, May 2026), and this view provided directional guidance for the brokerage industry's broader embrace of AI trading. IG's strategy of starting with read-only access reflects the cautious approach taken by traditional regulated brokers as they seek a balance between compliance requirements and technological innovation.
The AI transformation of the fintech sector is also accelerating. According to reports, Revolut built a trading platform prototype using AI in only about 30 minutes, triggering industry discussion over whether AI may gradually replace traditional broker trading interfaces. As the MCP ecosystem continues to expand, brokers are no longer facing the question of "whether to integrate AI," but a strategic contest over "how deeply and at what pace to integrate it."
Common Questions About Broker AI Integration Trends
What is the fundamental difference between the MCP protocol and traditional trading APIs?
Traditional trading APIs require developers to write code to call specific interfaces, creating a relatively high technical threshold. The MCP protocol allows AI assistants to interact directly with trading platforms through natural language, enabling users to access functionality without programming skills. In addition, MCP provides a standardized secure connection mechanism, making integration methods more unified across different AI platforms and brokers.
Why did IG choose a read-only model instead of directly opening AI trade execution permissions?
As a traditional broker regulated by ASIC, IG bears direct compliance responsibility for retail client trading safety. The read-only model provides AI-assisted functionality without introducing trade execution risk, making it a strategy that balances technological innovation with prudent compliance. Compared with platforms such as eToro and Bybit, IG faces stricter regulatory constraints. A gradual approach helps it expand permissions step by step after regulatory acceptance.
What is the key functional difference between eToro's Agent Portfolios and IG's CFD assistant?
eToro's Agent Portfolios allows AI agents to execute actual trades within independent sub-portfolios. Users can set capital limits and strategy boundaries, while the AI has the ability to place and close orders. IG's CFD assistant is strictly limited to read-only queries, and the AI cannot execute any trading instructions. The two represent different stages in brokers' AI transformation paths: the former is AI-driven trade execution, while the latter is AI-assisted information access.
What challenges may arise from large-scale adoption of the MCP protocol in financial trading?
The main challenges cover four levels: legal responsibility for AI trading decisions remains unclear; agentic trading may amplify market volatility and systemic risk; regulatory standards vary significantly across countries and regions, making cross-border compliance more difficult; and there is an inherent tension between compliance auditing and explainability requirements for AI model outputs and the black-box nature of such models. The current divergence in AI permission settings across platforms largely reflects the direct impact of these uncertainties on business decisions.
