Bybit has taken a step beyond standard exchange-side AI features and moved closer to what it describes as AI-native trading infrastructure. The crypto exchange said it has launched its official Model Context Protocol, or MCP.
This gives professional traders and developers a standardized way to build multi-agent trading systems that can connect directly with Bybit’s market, trading, account, and data-streaming functions.
The launch matters because it changes how traders can interact with an exchange. Instead of relying on custom API wiring, manual integrations, or fragmented automation setups, Bybit’s MCP is designed to let users build AI-driven trading desks that work through natural language commands.
In practice, that means traders can use popular AI agents such as Claude, ChatGPT, Cursor, and other compatible platforms to pull live market data, execute orders, monitor portfolios, and manage workflows without having to constantly manage the plumbing behind the scenes.
Bybit said the new layer is meant to support both single-agent and multi-agent architecture, which is where the bigger shift appears to be happening. A single AI assistant can already help automate parts of a trading workflow, but multi-agent systems open the door to more advanced setups where different AI agents handle different tasks at the same time.
For instance, one AI agent could monitor crypto price action, another could check account exposure, while another handles execution logic or risk management for the user. That kind of orchestration is what Bybit says its MCP is built to support.
Enhancing AI Trading Infrastructure
The company is framing the release as an infrastructure upgrade rather than a consumer-facing feature. That distinction is important. In recent years, many exchanges and fintech platforms have rushed to add AI labels to existing tools, often using them as chat assistants or help bots. Bybit’s pitch is different.
It wants to position itself as the plumbing layer that allows AI systems to discover and use exchange tools through a common protocol. That approach leans into interoperability, making it easier for AI applications that support MCP to connect with Bybit without each team building a separate integration from scratch.
According to the company, the protocol unlocks four core modules that traders need. The market data layer gives access to real-time ticker feeds, candlestick data, order book snapshots, and fee schedules.
The trading module, which Bybit says will continue to expand, is designed to support spot markets, perpetual futures, conditional orders, stop-loss and take-profit orders, and leveraged positions.
The account and asset module provides visibility into balances, open positions, order history, and transaction records. A real-time WebSocket data stream is also included, allowing persistent updates on market movement, execution activity, and position changes for latency-sensitive strategies.
Victor Wu, head of AI Agent Architecture at Bybit, said the company is not just building AI features, but AI infrastructure. He described the launch as a shift from standalone tools to a standardized protocol layer that can support multi-agent orchestration and more sophisticated automation.
He added, “We have moved the needle in AI innovation from developing standalone tools to creating a standardized protocol layer. Bybit’s MCP enables multi-agent orchestration and sophisticated automation. This is the future of trading infrastructure when AI becomes the primary interface.”
Building the Future of Trading
Security is another area that Bybit is emphasizing heavily. The exchange says the architecture is built around credential isolation, meaning it does not store, transmit, or access API keys or private keys used by agents.
Instead, API keys remain on the developer infrastructure, while Bybit receives authenticated requests from the MCP implementation. The company also says permissions can be scoped in a granular way, allowing keys to be limited to read-only access, specific trading permissions, or full execution capabilities.
It has also highlighted testnet validation, which gives users a way to test strategies with simulated funds before putting real capital at risk. The MCP setup runs in user-controlled environments, which, Bybit says, removes third-party intermediaries from the credential chain.
Bybit’s MCP is available immediately for integration and testing, with technical documentation, code samples, and deployment guidance already public. The company has pointed developers to a GitHub repository, a quick-start command, and package documentation.
It also noted compatibility with Claude Desktop, ChatGPT, Cursor, VS Code, and other MCP-compatible applications. Founded in 2018, Bybit says it now serves more than 80 million users worldwide and remains focused on bridging traditional finance and decentralized finance through secure custody, intuitive tools, and blockchain infrastructure.
With the MCP release, the exchange is signaling that its next phase may be less about adding AI on top of its platform and more about rebuilding the trading stack around AI from the ground up.
