News
·
Pyth Pro MCP Server 1.1
Pyth Pro MCP Server 1.1: empowering AI agents with direct access to market data for research, seamless one-time key setup, and tick-level historical data.
Financial markets are becoming software-driven
Financial markets are moving toward continuous, automated workflows. Tokenized assets, real-time risk systems, algorithmic research, and AI agents all depend on market data that can be retrieved programmatically and understood immediately.
A recent Forbes analysis puts the scale of this shift into context. The tokenized asset market, valued at roughly $600 billion today, is projected to reach nearly $19 trillion by 2033, according to BCG and Ripple.
The market for agentic AI is projected to grow from $5.2 billion in 2024 to nearly $197 billion by 2034, representing a compound annual growth rate of about 44%.
These figures point to a change in how financial data will be consumed. As agents move from isolated experiments into research, risk, monitoring, and decision-making workflows, they will need to query more markets, more frequently, and with less human intervention.
An agent monitoring an earnings announcement may need to find the right feed, retrieve the price at the moment of the event, inspect every update that followed, and compare the result with another asset. It needs structured data that is current, historical, cross-asset, and precise enough to support the analysis.
Pyth Pro MCP is built for this workflow. It gives agents a direct way to discover feeds and retrieve the market observations behind a financial answer. That is the reason Pyth is improving access to market data for AI agents.
Agents need market state, not just market commentary
AI agents can already search the web, read filings, summarise earnings, and identify the news surrounding a market event. That gives them financial information. It does not necessarily tell them what the market did next.
A useful financial answer may require the current price, the price at a specific timestamp, the movement immediately after an announcement, or the relationship between several assets over the same period. Those observations need to come from structured market data rather than articles, search results, or secondary commentary.
This is the difference between market narrative and market state.
Pyth Pro provides current prices, historical observations, candlestick data, funding rates, market sessions, and high-frequency update channels across asset classes. Through MCP, AI assistants can access those tools inside the same workflow where they are researching and reasoning.
That is also the idea behind the work Pyth and Exa are exploring. Exa provides agentic search and research infrastructure, while Pyth provides structured market data.
Exa can help an agent identify an earnings announcement, research the companies involved, and synthesise the relevant information. Pyth can provide the current and historical prices, candlestick data, and market observations needed to analyse the response.
The aim is to give agents the right source for each part of a financial question.
What’s new in Pyth Pro MCP Server 1.1
Pyth Pro MCP Server 1.1 makes access simpler, improves feed discovery, and gives agents more detailed market data.
Set your key once. Users can now add their Pyth Pro access token to their MCP client configuration instead of passing it with every request. For Claude Code, users can run:
The token belongs to the user and is sent through the client configuration. The hosted server does not use a shared Pyth Pro key, and users can omit the access_token parameter from individual calls when their client is already configured.
See how a price moved around an event. The new get_price_range tool returns every available price update in a window of up to 60 seconds. An agent can examine the sequence of updates around an earnings announcement, liquidation, policy decision, or other market event instead of comparing only two snapshots.
This gives agents a more precise view of short-term price behaviour and helps them explain the path a market took, not only its starting and ending points.
Know what you can access. When a user provides a Pyth Pro key, get_symbols shows the feeds available to Pro keys and flags which ones the user’s plan covers. An assistant can distinguish between a feed that exists, a feed that is outside the user’s plan, and a feed that is not currently live before requesting a price.
Use familiar pairs. Users can ask for BTC/USD without looking up the full Pyth symbol first. The same approach works across crypto, FX, equities, and metals. The server resolves the pair against the available catalog and reports when more than one possible match remains.
Work with more market data. The server now exposes funding rates, EMA prices, market-session information, update channels down to 50 milliseconds, interest-rate feeds, and NAV feeds. This gives agents a broader view of market conditions across crypto, equities, FX, commodities, rates, and other financial instruments.
Get clearer answers. When a feed is unavailable, the response explains why. The reason may be an invalid token, an entitlement gap, an inactive feed, a missing price, or an unsupported channel. The assistant can then explain the limitation instead of returning a generic error.
From financial research to market analysis
Consider a question such as:
What happened to a company after its latest earnings announcement, and how did its stock perform over the following five sessions?
Exa can identify the announcement, retrieve relevant filings and reporting, and provide a detailed analysis of the event. Pyth can add the corresponding price observations and candlestick data for the requested period. The agent can then connect the event with the market response in one structured answer.
This is the direction Pyth and Exa are testing: combining web research with the market data that describes what happened in the market itself. The work is still in an early phase, but the use case is clear. Agents need the right source for each part of a financial question.
Pyth Pro MCP Server 1.1 makes that access more direct. Users can configure a key once, discover the feeds they can use, retrieve market data at higher resolution, and understand why a requested feed may be unavailable.
Explore the Pyth Pro MCP documentation or run the server locally.
