Skip to main content
Glama
pythia-oracle-mcp

pythia-oracle-mcp

Official

Pythia Oracle MCP Server

PyPI License: MIT pythia-oracle-mcp MCP server

Every smart contract deserves intelligence, not just data.

Pythia is the first oracle delivering calculated technical indicators on-chain — EMA, RSI, VWAP, Bollinger Bands, volatility — for any token, on any Chainlink-supported chain. The same indicators traders use, available to smart contracts and AI agents with a single call via Chainlink.

Pythia Events lets smart contracts subscribe to indicator conditions (RSI below 30, EMA crossover, Bollinger breakout) and get called automatically when they trigger. No keeper, no off-chain bot, no polling — your contract reacts to markets on its own.

Pythia Visions delivers walk-forward validated market intelligence on-chain — pattern type, confidence score, indicator snapshot, and feeds-to-watch for confirmation, all in one event. Backtested across 9 years of history. FREE to subscribe. For the current live token + pattern catalog with accuracy stats and fire frequency, call get_visions_info from the MCP tools below.

Why Pythia?

Most oracles only give you price. Pythia gives you computed analysis: EMA, RSI, Bollinger Bands, VWAP, volatility — for tokens like BTC, SOL, TAO, RENDER, ONDO, AAVE, UNI, and more, across 4 timeframes, delivered on-chain via Chainlink. New tokens and indicators are added on demand. If your AI agent, DeFi protocol, or trading bot needs on-chain RSI, EMA, or Bollinger Bands — Pythia is the only source.

Use cases:

  • AI trading agents that need on-chain technical signals

  • DeFi vault rebalancing based on RSI or volatility thresholds

  • Smart contract risk management using Bollinger Band width

  • AI-powered portfolio analysis with real-time calculated metrics

  • Event-driven strategies — subscribe to RSI thresholds or EMA crossovers, your contract gets triggered automatically

  • Automated DeFi bots without keepers — no Gelato, no cron jobs, no off-chain infrastructure

Related MCP server: Crypto Indicators MCP Server

Quick Start

pip install pythia-oracle-mcp

Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "pythia-oracle": {
      "command": "pythia-oracle-mcp"
    }
  }
}

Claude Code

claude mcp add pythia-oracle -- pythia-oracle-mcp

Cursor / Windsurf / VS Code

Add to MCP settings:

{
  "pythia-oracle": {
    "command": "pythia-oracle-mcp"
  }
}

OpenAI Agents / GPT

Any MCP-compatible client works — just point it at pythia-oracle-mcp.

Run directly

python -m pythia_oracle_mcp

Available Tools

Tool

Description

list_tokens

All tracked tokens with status, uptime, and data sources

get_token_feeds

All indicator feed names for a specific token

get_market_summary

System-wide overview — tokens by status, ecosystem coverage, infrastructure health

check_oracle_health

Per-token 30-day uptime (worst-first), data source status, incident report

get_contracts

All contract addresses (operator, consumers, faucet, LINK)

get_pricing

Pricing tiers and when to use each one

get_integration_guide

Ready-to-deploy Solidity code for any tier

get_events_info

How Pythia Events work — subscribe to indicator conditions, get triggered on-chain

get_events_guide

Solidity code and deployment steps for event subscriptions

subscribe_info

Subscription details — conditions, pricing, refund mechanics

get_visions_info

Pythia Visions overview — walk-forward validated patterns, fire-frequency disclosure, contract address

get_visions_guide

Solidity code to subscribe to Visions and listen for VisionFired events

get_vision_history

Recent Visions fired for a token with pattern breakdown and confidence stats

get_vision_payload

Full enriched object for a fired Vision — failure profile, cooldown context, concurrent fires

lookup_event_feed

Reverse-lookup an Event feedId (bytes32) to its human-readable feed name

list_subscriptions

Enumerate active Pythia Event subscriptions for an owner address

get_feed_value

Latest computed value for any Pythia indicator feed (off-chain cache)

get_indicator_history

Read the free, immutable public history for a feed over a date range (≤31 days/call) — per-day coverage + optional ABOVE/BELOW verdict (TRUE / FALSE / INSUFFICIENT_DATA) with the exact day-file URLs so anyone can re-verify

Example Prompts

Ask your AI agent:

"What indicators does Pythia have for Bitcoin?"

Calls get_token_feeds("bitcoin") — returns all indicator feeds for Bitcoin, grouped by type.

"Is Pythia reliable enough to integrate?"

Calls check_oracle_health() — returns per-token uptime, data source health, and active incidents.

"Give me a Solidity contract to consume Pythia's speed bundle"

Calls get_integration_guide("speed") — returns a complete, deployable contract with correct addresses and job IDs.

"What tokens does Pythia cover and are they all working?"

Calls get_market_summary() — returns ecosystem coverage, status breakdown, and infrastructure health.

"How do Pythia Events work? I want my contract to react when BTC RSI drops below 30."

Calls get_events_info() — returns how subscriptions work, supported conditions, and pricing.

"Give me the Solidity code to subscribe to an EMA crossover event."

Calls get_events_guide() — returns a deployable EventSubscriber contract with subscribe/receive pattern.

"What are Pythia Visions? What patterns does it detect?"

Calls get_visions_info() — returns walk-forward validated patterns with accuracy range, fire frequency, contract address, and how it works.

"Show me recent BTC Visions that fired"

Calls get_vision_history("BTC") — returns recent pattern detections with confidence and price.

"Give me Solidity code to subscribe to Pythia Visions"

Calls get_visions_guide() — returns a contract that subscribes to VisionFired events.

What Pythia Provides

  • Any token, any Chainlink-supported chain — currently serving BTC, SOL, TAO, RENDER, ONDO, AAVE, UNI, MORPHO, and more, with new tokens added on demand

  • 6 indicator types: EMA, RSI, Bollinger Bands (upper/lower), VWAP, Volatility, USD Price

  • 4 timeframes: 5-minute, 1-hour, 1-day, 1-week

  • 4 pricing tiers: Discovery / Analysis / Speed / Complete — for current LINK fees call get_pricing

  • Free trial: PythiaFaucet contract — no LINK needed

  • Pythia Events: Subscribe to indicator conditions (ABOVE/BELOW thresholds) — your contract gets called when they trigger. Prepaid in LINK, unused time refunded on cancel or fire. No keeper infrastructure needed.

  • Pythia Visions: Walk-forward validated market intelligence on-chain — pattern type + confidence + indicator snapshot + feeds-to-watch, delivered via Chainlink. FREE to subscribe. For the live list of patterns + tokens, call get_visions_info.

  • Public indicator history: free, immutable JSON archive of every published indicator value — one file per feed per closed UTC day at https://pythia.c3x-solutions.com/history/{chain}/{feed_name}/{YYYY-MM-DD}.json (start from history/manifest.json). Re-derive or audit any indicator condition over a date range without an oracle call; files never change once published.

Data Freshness

This MCP server is a thin client over https://pythia.c3x-solutions.com/feed-status.json, the live status feed updated every 15 minutes by the data engine. New tokens, new patterns, contract addresses, and pricing changes appear in MCP tools within seconds of the next data-engine cycle — no package update required.

If the live feed is unreachable, MCP tools raise a clear error rather than serve stale baked-in data. Retry shortly or check status. (As of v0.9.0, fail-loud — earlier versions had baked-in fallbacks that drifted from production state and have been removed.)

Integration Examples

See pythia-oracle-examples for Solidity contracts with Hardhat setup — ready to deploy on any Chainlink-supported network.

License

MIT

Available Tools

18 tools
check_oracle_healthA

Check the reliability and uptime of Pythia's oracle system.

Returns per-token 30-day uptime (sorted worst-first so problems surface immediately), recent daily status history, data source health, and infrastructure status. Use this to verify Pythia's reliability before integrating or relying on its data.

Inputs: no named parameters. Returns: result

get_contractsA

Get Pythia contract addresses for on-chain integration. Shows all supported chains.

Inputs: no named parameters. Returns: result

get_events_guideA

Get Solidity code to subscribe to Pythia Events (indicator alerts).

Returns a complete contract that approves LINK, subscribes to an indicator alert, listens for PythiaEvent, and can cancel for a refund.

Inputs: no named parameters. Returns: result

get_events_infoA

Get overview of Pythia Events — on-chain indicator alert subscriptions.

Returns pricing, supported conditions, subscriber flow, registry addresses per chain, and current subscription stats. Events let you subscribe once and get notified when an indicator crosses a threshold.

Inputs: no named parameters. Returns: result

get_feed_valueA

Get the latest computed value of a Pythia indicator feed.

Reads from the live cache (feed_values table) populated by the indicator pipeline on every cycle. Off-chain AI agents use this when reasoning about a Vision context, choosing an Event threshold, or sanity-checking a feed's current level. On-chain consumers should request the value through oracle.request() to get a Chainlink-attested response — see get_integration_guide().

Args: feed_name: full feed name (e.g. 'bitcoin_RSI_1H_14', 'pol_EMA_5M_20').

Returns: Latest value + computed_at + chain, one block per chain if a feed exists on multiple chains. If the feed has no cached value, returns a diagnostic pointer (warm-up window, deactivated, or unknown name).

Inputs: feed_name (required). Returns: result

get_indicator_historyA

Read Pythia's public, immutable indicator history for a feed over a date range — and optionally settle a condition against it.

Backed by the free day-file archive at https://pythia.c3x-solutions.com/history/{chain}/{feed_name}/{YYYY-MM-DD}.json (one file per feed per closed UTC day, 5-minute points, never rewritten once published). Use it to audit or re-derive "was RSI below 30 at any point last week?", to reconstruct what a Vision or Event saw, to backtest a threshold before subscribing to an Event, or to settle a prediction-market style question from public inputs anyone can re-fetch and verify.

Args: feed_name: full feed name, same key as Feeds/Events (e.g. 'bitcoin_RSI_1D_14'). start_date: first UTC day, YYYY-MM-DD (inclusive). end_date: last UTC day, YYYY-MM-DD (inclusive). Defaults to start_date. At most 31 days per call — split longer ranges. chain: delivery chain the feed is archived for (e.g. 'polygon'). Only needed when the manifest lists the feed on more than one chain. condition: optional 'ABOVE' or 'BELOW' — with threshold, evaluates whether ANY point in the range satisfies it. threshold: numeric threshold for condition, in the feed's own units.

Returns: Per-day coverage (points, min, max, first, last), the range summary, the day-file URLs used (so a verifier can re-fetch the exact inputs), and — when condition+threshold are given — a verdict: TRUE (first matching timestamp + value), FALSE (full coverage, no match), or INSUFFICIENT_DATA (missing day-files or gaps large enough that FALSE cannot be asserted). Never interpolates across gaps.

Inputs: feed_name (required), start_date (required), chain, end_date, +2 more. Returns: result

get_integration_guideA

Get Solidity code to integrate Pythia into a smart contract.

Args: tier: 'discovery' (single value), 'analysis', 'speed', or 'complete'. chain: Optional chain key (e.g. 'mainnet', 'amoy', 'arbitrum'). When unset, the embedded Solidity uses the first available chain's addresses (Polygon mainnet by convention) and the prose lists every chain where this tier's consumer is deployed.

Inputs: tier, chain. Returns: result

get_market_summaryA

Get a summary of all tokens tracked by Pythia with operational overview.

Returns system-wide stats, tokens grouped by status, uptime distribution, data source health, and infrastructure status. Useful for quickly understanding what Pythia covers and whether the system is healthy.

Inputs: no named parameters. Returns: result

get_pricingA

Get Pythia pricing tiers and free trial info. Prices are live from the data feed.

Inputs: no named parameters. Returns: result

get_token_feedsA

Get all available indicator feeds for a specific token.

Shows every feed name (EMA, RSI, Bollinger, Volatility across all timeframes), the token's reliability stats, and data source count. Feed names are what you pass to the on-chain oracle to request data.

Args: engine_id: Token engine ID (e.g., 'bitcoin', 'solana', 'bittensor', 'aave', 'pol'). Use list_tokens() to see all available IDs.

Inputs: engine_id (required). Returns: result

get_vision_historyA

Get recent Pythia Visions fired for a token with pattern breakdown and stats.

Args: token: Token symbol to check (default: BTC). Case-insensitive. Currently live: BTC, ETH.

Inputs: token. Returns: result

get_vision_payloadA

Get the full enriched object for a fired Pythia Vision by id.

Returns the rich AI-facing companion to the on-chain VisionFired event: pattern metadata with numeric ranges, failure profile (avg return when correct, avg drawdown when wrong, worst drawdown), cooldown context (hours since last same-pattern fire on this token, confidence delta vs last fire), and concurrent fires from other tokens within the last 24h.

Lightweight on-chain consumers can decode the VisionFired payload bytes directly. AI agents reasoning about a specific Vision should use this tool — it contains the data needed to size positions and compare against historical failure modes, which the on-chain event payload does not.

Args: vision_id: integer id of the Vision (returned by get_vision_history)

Returns: Multi-section text report. If vision_id is not in the recent window (last 20 fires per token), returns a helpful pointer to history.

Inputs: vision_id (required). Returns: result

get_visions_guideA

Get Solidity code to subscribe to Pythia Visions and listen for VisionFired events.

Returns a complete contract that subscribes to the PythiaVisionRegistry, receives VisionFired events with pattern type, confidence, direction, price, and full analysis payload. Subscription is FREE (no LINK required).

Inputs: no named parameters. Returns: result

get_visions_infoA

Get overview of Pythia Visions — walk-forward validated market intelligence on-chain.

Returns the walk-forward validated patterns with accuracy stats, the Vision Registry contract address, subscription info (FREE), evaluation frequency, and supported tokens. Visions are pattern detections that passed walk-forward validation across multiple years of history. Live token + pattern set is returned in the response (canonical source: feed-status.json visions section).

Inputs: no named parameters. Returns: result

list_subscriptionsA

Enumerate active Pythia Event subscriptions owned by an address.

Returns every subscription where active=true (not yet fired, expired, or cancelled). Without this tool, dApps and dashboards have to replay every SubscriptionCreated log from the registry deploy block to discover what an owner is currently subscribed to.

Args: owner_address: subscriber wallet address ('0x...', case-insensitive).

Returns: Multi-section report listing each active subscription with feed name, condition + threshold, expiry, registry address, and creation tx.

Inputs: owner_address (required). Returns: result

list_tokensA

List all tokens tracked by Pythia with status and reliability info.

Returns token symbols, categories, data source count, 30-day uptime, and operational status. Covers cross-chain tokens (BTC, SOL, TAO, RENDER, ONDO, etc.) and DeFi tokens.

Inputs: no named parameters. Returns: result

lookup_event_feedA

Reverse-lookup a Pythia Event feedId (bytes32) to its human-readable feed name.

Subscribers receive bytes32 feedId in SubscriptionCreated and PythiaEvent events. This tool maps that hash back to the canonical feed name (e.g. 'pol_RSI_5M_14') so dApps don't need to maintain their own feedId → name table or query the registry contract on every event.

Args: feed_id_hex: bytes32 hash, with or without '0x' prefix, any case.

Returns: Single-section report with feed_name + matching token + indicator suffix. If the hash is not in the registered lookup table, returns a diagnostic pointer.

Inputs: feed_id_hex (required). Returns: result

subscribe_infoA

Plan a specific Pythia Events subscription with cost and exact calls.

Args: feed_name: Feed name to monitor (e.g. 'pol_RSI_5M_14', 'bitcoin_EMA_1H_20') condition: 0=ABOVE, 1=BELOW, 2=CROSSES_ABOVE, 3=CROSSES_BELOW days: Subscription duration in days (1-365)

Inputs: feed_name (required), days, condition. Returns: result

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.8/5.0

    Scored across 18 tools

    Disambiguation2/5

    Several tools have unclear boundaries: list_tokens, get_market_summary, and check_oracle_health all return uptime, status, and data-source health, with list_tokens already including 30-day uptime. Pricing/cost info also overlaps across get_pricing, get_events_info, and subscribe_info. Most other tools are distinct, but these multi-way overlaps make tool selection genuinely ambiguous.

    Naming Consistency4/5

    The vast majority of tools follow a clean lowercase snake_case verb_noun pattern, and the get_* family is dominant. Parallel names like get_events_info/get_events_guide and get_visions_info/get_visions_guide reinforce consistency, though a few outliers such as subscribe_info and check_oracle_health use different verbs.

    Tool Count3/5

    18 tools sits in the 16–25 range that feels heavy for a single-purpose server. The domain has multiple sub-areas, but the overlapping status/health tools suggest the surface could be consolidated. It is not excessive enough for a 2, but it is beyond the ideal 3–15 scope.

    Completeness4/5

    The tool set covers token discovery, feed data, historical verification, event subscription planning, vision analysis, contract addresses, integration guides, and pricing, which supports most real workflows. Minor gaps exist, such as no direct historical Event fire enumeration and no on-chain subscription actions, but agents can work around these via guides and get_indicator_history.

    Maintenance

    ActivityStale
    ResponsivenessOutdated data

    Related MCP Connectors

    Related MCP Servers