Pyth Network has released version 1.1 of its Pyth Pro MCP Server, expanding the structured financial data that compatible AI assistants can query directly during research and analysis. The update adds higher-resolution historical pricing, simpler feed discovery and broader access to market information across crypto, equities, foreign exchange, metals, rates and net asset values.
According to Pyth’s official release announcement, the Model Context Protocol server connects AI clients with Pyth Pro data through a hosted endpoint. Users provide their own Pyth Pro access key, which determines which feeds they can query. The server itself does not carry a shared Pyth credential or give an AI system unrestricted access to the full data catalog.
Version 1.1 also simplifies authentication by allowing the access token to be configured once inside a compatible MCP client rather than passed with every individual request. That setup is currently documented for environments including Claude Code, Claude Desktop, Cursor and VS Code.
Tick-Level Data Adds Event Analysis
One of the most significant additions is get_price_range, which returns every available price update within a window of up to 60 seconds. An AI assistant can use that sequence to examine how an asset moved immediately around an earnings release, liquidation, policy announcement or other short-duration market event instead of comparing only two price snapshots.
The feature should not be interpreted as unlimited tick-history retrieval. Pyth’s documentation restricts each range request to a 60-second interval, while broader historical analysis can use point-in-time observations or OHLC candlestick data. The server also supports update channels as fast as 50 milliseconds where the underlying feed permits them.
Feed discovery has become more explicit as well. The get_symbols tool can show which markets exist, whether a feed is active and whether the user’s subscription entitles them to retrieve its data. That reduces a common failure mode for agents by distinguishing an unavailable market from a valid feed that simply sits outside the user’s plan.
Users can also request familiar identifiers such as BTC/USD instead of first locating Pyth’s full internal symbol. Pyth says pair-based resolution currently works across crypto, FX, equities and metals, while the broader catalog includes funding rates, interest rates, NAVs, commodities and other financial data.
MCP Moves Financial Data Into Agent Workflows
The release places Pyth inside a broader movement toward standardized interfaces between AI systems and financial infrastructure. Aave recently launched an MCP server for AI-assisted DeFi workflows, while Base MCP connects assistants with wallet and application workflows. Pyth’s implementation occupies a different layer because its documented role is data retrieval rather than transaction preparation or asset execution.
That distinction becomes more important as MCP is also used for direct trading systems. Gemini, for example, has connected the protocol to agentic crypto execution through its trading infrastructure. Pyth Pro MCP supplies market state that could inform those types of workflows, but the current server does not itself place orders, sign transactions or control capital.
Pyth has already extended its data infrastructure beyond DeFi-native use cases. Its Pro feeds serve as market data for Kalshi’s commodity contracts, demonstrating that the same underlying data layer can support regulated markets as well as AI research tools.
Pyth Pro MCP Server 1.1 therefore expands what an AI assistant can know about financial markets rather than what it can autonomously execute. The measurable development is a richer machine-readable interface for current prices, historical observations, market sessions, funding data and other structured feeds; whether that becomes a widely used financial-agent data layer will depend on actual developer integrations and recurring query activity.