Pythia Oracle MCP

What pythia-the-oracle/pythia-oracle-mcp MCP server does

The pythia-the-oracle/pythia-oracle-mcp MCP server gives AI agents access to Pythia’s calculated crypto market data and integration documentation. Pythia provides EMA, RSI, Bollinger Bands, VWAP, volatility, and USD price feeds for tracked tokens. The published coverage includes 22 tokens, 484 indicator feeds, and four timeframes: 5-minute, 1-hour, 1-day, and 1-week.

The tools support both inspection and implementation planning. An agent can identify supported tokens and feeds, read the latest value for a feed, review oracle uptime and incidents, or retrieve contract addresses. It can also obtain pricing information and Solidity examples for consuming Pythia feeds, subscribing to Events, or listening for Visions.

How it works

The server is a thin client over Pythia’s live feed-status service at https://pythia.c3x-solutions.com/feed-status.json. That source is updated every 15 minutes. Changes such as newly supported tokens, patterns, contract addresses, and pricing are therefore exposed through the MCP tools without requiring a package update.

Available tools cover several related data paths:

• Token and infrastructure discovery through list_tokens, get_token_feeds, and get_market_summary.
• Reliability checks through check_oracle_health, including 30-day uptime, data-source status, and incidents.
• Integration details through get_contracts, get_pricing, and get_integration_guide.
• Event subscriptions that react to indicator conditions such as threshold crossings or EMA crossovers.
• Visions data containing pattern detections, confidence, indicator snapshots, and related feed information.
• Historical verification through get_indicator_history, which reads immutable daily JSON archives for up to 31 days per call.

The history tool can return daily coverage and optional ABOVE or BELOW verdicts with TRUE, FALSE, or INSUFFICIENT_DATA results. It also provides the underlying day-file URLs for independent verification.

Setup and configuration

Install the Python package with:

pip install pythia-oracle-mcp

Run the server with the installed pythia-oracle-mcp command, or use python -m pythia_oracle_mcp. MCP client configuration uses a local stdio process. The README provides configuration examples for Claude Desktop, Claude Code, Cursor, Windsurf, and other MCP-compatible clients.

No environment variables or API credentials are specified in the provided setup instructions. The server reads its operational data from the public Pythia endpoint rather than from values baked into the package.

Tools and capabilities

The pythia-the-oracle/pythia-oracle-mcp MCP server exposes tools for current data, historical data, and on-chain integration:

• Discover supported tokens, indicator feeds, ecosystem coverage, and infrastructure status.
• Retrieve the latest computed value from an off-chain feed cache.
• Inspect contract addresses for the operator, consumers, faucet, and LINK token.
• Read pricing tiers and obtain ready-to-deploy Solidity integration code.
• Learn how Pythia Events work, including subscription conditions, pricing, refunds, and deployment steps.
• Query recent Vision firings and retrieve detailed payloads with failure profiles, cooldown context, and concurrent fires.
• Reverse-lookup Event feed IDs and list active subscriptions for an owner address.
• Audit published indicator history without making an oracle call.

Pythia Events are designed for contracts that should react when an indicator condition occurs. Pythia Visions provide walk-forward-validated pattern information on-chain and are free to subscribe to according to the supplied material.

Limitations and notes

The MCP server depends on the live Pythia data service. If that feed is unreachable, tools return an error instead of serving stale fallback data, so callers should retry or check service status. Historical requests are limited to 31 days per call.

The repository describes Pythia’s on-chain products and integration examples, but the MCP tools themselves provide information and data access rather than deploying contracts. Pricing, supported assets, contract addresses, and live pattern catalogs should be retrieved through the relevant tools because they can change independently of the package release.