Bybit Introduces MCP Protocol to Support AI-Driven Multi-Agent Trading Systems

A shift toward AI-native trading infrastructure

Bybit has introduced a new infrastructure layer designed to support the growing role of artificial intelligence in trading. The company’s Model Context Protocol (MCP) provides a standardized way for AI agents to interact with trading systems, marking a move away from tools built primarily for human users.

The protocol allows compatibility with widely used AI platforms, including Claude, ChatGPT, and development environments like Cursor and VS Code. Instead of requiring direct API integrations, traders and developers can deploy automated strategies using natural language inputs, simplifying the process of building and managing trading systems.

How It May Look on Claude – example:

This approach reflects a broader shift in the industry, where trading infrastructure is increasingly expected to support continuous, automated decision-making by AI systems rather than manual execution.

Bybit’s MCP is designed to scale from single-agent setups to more complex multi-agent architectures. These systems can coordinate tasks such as analyzing market conditions, executing trades, and adjusting portfolio positions in real time.

Core modules and system capabilities

The MCP framework is built around four primary modules that cover essential trading functions. These include access to live market data, trading execution tools, account management features, and real-time data streams via WebSocket connections. Together, they allow AI agents to operate across the full trading lifecycle, from data analysis to execution.

For example, agents can monitor price movements through ticker data and order book snapshots, place trades across spot and derivatives markets, and track portfolio balances and transaction history. The addition of persistent data streams supports strategies that rely on low-latency updates and rapid reaction times.

A visual example on page 2 of the document shows how such a system might function within an AI interface, where a user prompts an agent to retrieve a Bitcoin order book snapshot in real time, demonstrating how natural language commands translate into actionable trading insights.

Victor Wu, Head of AI Agent Architecture at Bybit, described the release as part of a broader transition in how trading systems are built. “Every exchange is building AI features. Bybit is building AI infrastructure,” he said, adding that the protocol is intended to support more advanced automation and coordination between multiple AI agents.

Security design and developer access

Alongside automation capabilities, the MCP framework emphasizes user control over sensitive data. The system is designed so that API credentials remain within the user’s environment, with Bybit only receiving authenticated requests rather than direct access to keys.

Developers can assign specific permissions to API keys, limiting them to read-only access or enabling full trading capabilities depending on the use case. The platform also supports testing strategies in a simulated environment before deploying them with real funds, reducing potential risks during development.

The infrastructure is available for immediate use, with supporting documentation, code samples, and deployment tools provided to facilitate integration. It is compatible with a range of AI applications that follow the MCP standard, allowing developers to experiment with different setups and configurations.

Documentation: https://www.npmjs.com/package/bybit-official-trading-server

The bottom line

Bybit’s MCP release signals a move toward infrastructure designed specifically for AI-driven trading. By standardizing how AI agents interact with trading systems and removing the need for complex integrations, the protocol lowers barriers for building automated strategies. As AI tools become more embedded in financial workflows, systems like MCP may shape how trading platforms evolve to support increasingly autonomous operations.