Built, shipped, operated
Cerebrus Pulse
A multi-protocol crypto-intelligence API that gives AI agents real-time market data — distributed as an MCP server, a Python SDK, and a LangChain toolkit, paid per query via x402 micropayments.
- API
- Crypto
- x402
- MCP
Live interface
Query it live.
Pick an asset and pull a real read from the Cerebrus Pulse API — current price, RSI, EMA trend across timeframes, straight from the live endpoint.
Real-time technicals for 50+ Hyperliquid perpetuals — click to pull a live read from the Cerebrus Pulse API.
Operating brief
The constraint.
AI agents that trade or reason about markets need real-time intelligence, but crypto data is fragmented across venues, expensive to aggregate, and rarely packaged in a form an agent can call directly. There was no clean, pay-as-you-go way to give an agent technical analysis, sentiment, and liquidation data over a single interface.
What the system does
Exposes 15 tools — multi-timeframe technicals (RSI, EMAs, Bollinger Bands, VWAP, regime), sentiment, funding rates, open interest, liquidation heatmaps, correlation, market stress, and more.
Covers 50+ Hyperliquid perpetuals across six timeframes with confluence scoring.
Ships as an MCP server (Claude / Cursor / Windsurf), a typed Python SDK, and a LangChain toolkit — one intelligence layer, three integration paths.
Charges per query with x402 micropayments; free tools (health, coin list) work with no wallet configured.
Architecture
How it moves.
One gateway, three client surfaces. Paid tools sit behind an x402 paywall; free tools are open.
- AI client (Claude · Cursor · LangChain)
- MCP server / Python SDK
- Cerebrus gateway
- Hyperliquid + Chainlink data
Paid tools gated by x402 (USDC on Base or Solana); free tools need no wallet.
Build ledger
What carries it.
- Interface
- MCP server, Python SDK (PyPI), LangChain toolkit, CLI
- Payments
- x402 micropayments · USDC on Base & Solana
- Data
- 50+ Hyperliquid perpetuals · Chainlink oracles
- Analysis
- RSI, EMAs, ATR, Bollinger Bands, VWAP · regime & confluence scoring
Current status
Delivered.
Live at cerebruspulse.xyz; published to PyPI as cerebrus-pulse-mcp and cerebrus-pulse-python.
15 endpoints across six timeframes, load-tested at a sub-200ms p50.
Agent-commerce layer: x402 payments, Google A2A with Agent Card discovery, and ACP synthesis on Virtuals Protocol.