Horus Flow Intelligence — Institutional Orderflow for AI Agents
LocalAbout Horus Flow Intelligence — Institutional Orderflow for AI Agents
Horus Flow Intelligence is a market-data MCP server for quantitative traders and AI trading agents. It combines order-book flow, derivatives and equity context, signal reliability, and Cortex assessments, while keeping its tools read-only.
Key features
- 01Live crypto order flow and book pressure
- 02Derivatives, liquidation and market-regime data
- 03Equity flow and US trading-session context
- 04Market scans, events and signal history
- 05Cortex contradictions and execution boundaries
Use cases
- 01Assessing order pressure before a trade
- 02Comparing crypto signals with equity flows
- 03Scanning markets for breakouts or surges
- 04Reviewing historical signal hit rates
MCP support
Provides a remote Streamable HTTP MCP server and a local stdio option. Its read-only tools expose market flow, intelligence, events, signal records and service status to MCP clients.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
We haven’t collected a package or remote endpoint for this server. Follow the instructions in the project README.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project README🦅 HORUS CORTEX & FLOW
The Cognitive Market Brain for Autonomous AI Trading Agents
Don't let your AI trading agents trade on lagging candlestick echoes. Give them the whale's eyes and an institutional cognitive mind.
🤖 The Zero-Human Decision Loop in Action
When you connect Claude Desktop, Cursor, or an autonomous LangChain / AutoGen agent to Horus Cortex, your agent stops hallucinating and begins executing with Wall Street risk rigor:
┌─────────────────────────────────────────────────────────────────────────────────────────────┐
│ Autonomous Quant Agent (Claude 3.5 Sonnet / Cursor IDE) 🤖 │
│ │
│ Prompt: "BTC is pumping. Should we execute a 3x breakout long entry?" │
│ │
│ AI Calls Tool: `get_horus_cortex()` │
│ ◀ Horus Cortex Returns: │
│ { │
│ "regime_state": "TRANSITION", │
│ "trust_score": 26.5, │
│ "action_policy": { │
│ "directive": "Deteriorating Transition: Capital Preservation Lock (0.0x)", │
│ "trend_allowed": false, │
│ "ignition_allowed": false │
│ }, │
│ "market_vitals": { "taker_ratio": 0.24, "global_ignition": 0.08 }, │
│ "active_contradictions": [ │
│ { "id": "BULLISH_PRICE_vs_HOLLOW_BREADTH", "severity": "MEDIUM" } │
│ ], │
│ "execution_boundaries": { │
│ "invalidation_support": 77052.0, │
│ "breakout_resistance": 78608.0 │
│ } │
│ } │
│ │
│ AI Final Output: │
│ "⛔ TRADE ABORTED — CAPITAL PRESERVATION ENFORCED. │
│ Although spot price is making local highs, Horus Cortex flags a severe divergence: │
│ 1. Taker Buy/Sell ratio collapsed to 0.24 (Aggressive seller absorption). │
│ 2. Global market breadth is dormant (0.08) indicating an isolated, hollow squeeze. │
│ 3. Cortex Trust Score is 26.5/100 with active trend locks enabled. │
│ Action: Standing aside. Re-evaluating only if price confirms above $78,608 with volume." │
└─────────────────────────────────────────────────────────────────────────────────────────────┘
🛑 Stop Predicting Candles. Start Measuring Gravity.
Retail traders rely on lagging indicators (RSI, MACD, Moving Averages) which calculate what has already happened.
Horus is predictive, not reactive:
- Level 1-2 Physics (The Eye): Measures sub-second limit orderbook depth, spoofed liquidity walls, and aggressive taker execution delta before candlesticks print.
- Level 3 Forensics (The Radar): Tracks Liquidation Cascades (Gravity Direction) and Spot vs. Futures Speculation Ratios.
- Level 4 Cortex (The Mind): Synthesizes 7 independent evidence families with a real-time Contradiction Engine to output machine-executable position sizing multipliers and USD invalidation boundaries.
🏗️ Architecture
graph TD
%% Styling
classDef crypto_stream fill:#F3BA2F,stroke:#000,color:#000,stroke-width:2px;
classDef equity_stream fill:#000,stroke:#09b533,color:#09b533,stroke-width:2px;
classDef compute fill:#1A1F36,stroke:#00D6FF,color:#fff,stroke-width:2px;
classDef level3 fill:#1A1F36,stroke:#FF6B00,color:#fff,stroke-width:2px;
classDef cortex fill:#4A154B,stroke:#FFD700,color:#fff,stroke-width:3px;
classDef mcp fill:#632CA6,stroke:#fff,color:#fff,stroke-width:2px;
classDef client fill:#FF3366,stroke:#fff,color:#fff,stroke-width:2px;
%% Ingestion
subgraph Data_Pipelines [Websocket Ingestion]
B[Binance WSS <br/> L1/L2 Book]:::crypto_stream
A[Alpaca WSS <br/> IEX Equities - sample]:::equity_stream
BF[Binance Futures <br/> OI / Funding / Liquidations]:::crypto_stream
end
%% Level 1-2 Engine
subgraph Core_Engine [Level 1-2: Microstructure Physics]
IC[Imbalance Calculator <br/> Bid/Ask Walls]:::compute
FC[Flow Calculator <br/> Tape Deltas]:::compute
BC[Behavioral Court <br/> Spoofing & Absorption Rules]:::compute
end
%% Level 3 Intelligence
subgraph L3_Engine [Level 3: Institutional Intelligence]
LH[Liquidation Heatmap <br/> Gravity Zones]:::level3
XF[Cross-Exchange Flow <br/> Spot ↔ Futures]:::level3
CI[Composite Intelligence <br/> 4-Layer Verdict]:::level3
end
%% Level 4 Cognitive Cortex
subgraph L4_Cortex [Level 4: The Sovereign Cognitive Cortex]
HC[Horus Cortex Symphony 3.0 <br/> 7 Forensic Families + Contradiction Engine]:::cortex
TS[Penalized Trust Score 0-100]:::compute
EB[Execution Boundaries USD]:::compute
AP[Action Policy Multipliers for Bots]:::compute
end
%% Output
subgraph Output_Layer [Data Delivery & Agent Integration]
MCP[AI Agent FastMCP Server <br/> 8 Sovereign Tools]:::mcp
API[FastAPI REST Gateway <br/> 13 Endpoints]:::client
Dash[Real-Time Portal <br/> Live Microstructure]:::client
end
B --> IC
A --> IC
B --> FC
A --> FC
BF --> LH
BF --> XF
IC --> BC
FC --> BC
LH --> CI
XF --> CI
BC --> CI
CI --> HC
HC --> TS
HC --> EB
HC --> AP
HC --> MCP
HC --> API
BC --> Dash
📡 Complete API Catalog
Level 1-2: Orderflow Microstructure
| Endpoint | Method | Description | Tier |
|---|---|---|---|
/v1/flow/crypto/{symbol} | GET | Order book imbalance (top 20 levels) and taker flow from Binance's public stream. Any TRADING spot pair | Free: all pairs |
/v1/flow/crypto/{symbol}/history | GET | Snapshots every 15s with price, deltas and ratios | Builder: 60 min · Pro: 24 h |
/v1/flow/equity/{symbol} | GET | US equity flow from the IEX feed (about 3% of US volume, a sample) | Builder |
/v1/flow/equity/macro-blocks | GET | SPY buy/sell skew and IEX prints of $200k or more | Free |
/v1/status | GET | Freshness and uptime of every data source (no key) | Public |
Level 3: Forensic Market Intelligence
| Endpoint | Method | Description | Tier |
|---|---|---|---|
/v1/intelligence/climate | GET | WiseMan BTC regime and health | Builder |
/v1/intelligence/ignitions | GET | Volatility breakout scanner across ~750 assets, refreshed every minute | Builder |
/v1/intelligence/liquidation-heatmap | GET | Crowd positioning vs its 24h median and estimated liquidation levels (10 majors) | Pro |
/v1/intelligence/cross-exchange-flow | GET | Futures/spot volume vs its 7-day median, premium, funding and OI across Binance, Bybit and OKX | Pro |
/v1/intelligence/composite | GET | 0-100 score from climate, ignition, positioning and futures/spot flow | Pro |
/v1/intelligence/market-intelligence | GET | All of the above for 10 majors in one call | Pro |
Every Level 3 document carries a method field with its exact formula. Endpoints return 503 data_stale instead of serving data older than their freshness limit.
Level 4: The Sovereign Cognitive Cortex
| Endpoint | Method | Description | Tier |
|---|---|---|---|
/v1/intelligence/cortex | GET | Regime state, trust score, contradictions, execution boundaries and action policy | Pro |
/v1/intelligence/maestro | GET | Backward-compatible alias for the Cortex endpoint | Pro |
Agent Layer
| Endpoint | Method | Description | Plan |
|---|---|---|---|
/v1/book/pressure | GET | The spot order book of 30 Binance pairs in one call (top-five bid share, bid/ask ratio, 30 s taker flow), with its measured link to the next minute | Free: 3 rows · Trader: 30 |
/v1/scan | GET | Rank about 250 Binance perpetuals in one call: buyers or sellers in control, volume surge, breakout candidates, crowded funding, or custom filters | Free: 3 rows · Builder: 100 |
/v1/track-record | GET | Live hit rate of every signal at 1/5/15/30 min, by signal, symbol and confidence | Public |
/v1/track-record/signals | GET | Point-in-time ledger: what Horus said, at what price, and what happened next (since Aug 2026) | Builder |
/v1/events | GET | Changes since your last check: liquidation cascades, regime changes, new US halts, and more | Builder |
/v1/events/stream | GET | The same events as Server-Sent Events | Builder |
/v1/webhooks | POST GET DELETE | Signed webhooks for any event type | Pro |
/v1/equity/session | GET | US session phase, holidays, early closes, next open/close, active halts | Public |
/v1/equity/halts | GET | All US trading halts published today, with plain-language reasons | Public |
/v1/cross-asset | GET | BTC order flow next to SPY, QQQ, COIN, MSTR and IBIT flow | Builder |
/v1/status | GET | Freshness and uptime of every data source | Public |
Every directional flow reading carries a reliability block with its measured hit rate. Add ?view=compact to /v1/flow/crypto/{symbol} for a token-light answer.
🤖 Connect Your AI Agent
Remote MCP server (recommended)
https://flow.horustek.pro/mcp (Streamable HTTP).
- RapidAPI subscribers: send your RapidAPI key as
X-RapidAPI-Key: <key>. Calls go through RapidAPI, so your RapidAPI plan and quota apply. Subscribe on RapidAPI. - Horus keys:
X-API-Key: <key>orAuthorization: Bearer <key>. - No key: free tier: order flow for any pair, the public track record, US session context, the first 3 scanner results and service status.
Claude Code:
claude mcp add --transport http horus-flow https://flow.horustek.pro/mcp --header "X-RapidAPI-Key: YOUR_RAPIDAPI_KEY"
Any MCP client that supports remote servers:
{
"mcpServers": {
"horus-flow": {
"type": "http",
"url": "https://flow.horustek.pro/mcp",
"headers": { "X-RapidAPI-Key": "YOUR_RAPIDAPI_KEY" }
}
}
}
Tools (all read-only, compact JSON with an output schema)
| Tool | What it answers | Plan |
|---|---|---|
get_book_pressure | What does the order book say right now across 30 majors, and how much has that told about the next minute | Free (3 rows) / Trader |
scan_market | Where are buyers, sellers or volume surging across about 250 markets right now | Free (3 rows) / Builder |
get_crypto_flow | Live order flow for one pair, with the signal's measured hit rate | Free |
get_flow_history | Is pressure building or fading over the last hour | Builder |
get_market_overview | Climate, Cortex and 10 majors in one call | Pro |
get_derivatives | Funding, OI across Binance/Bybit/OKX, crowd positioning, liquidation levels | Pro |
get_cortex | Regime, trust score, contradictions, invalidation and breakout levels | Pro |
get_ignitions | Breakout scanner across ~750 assets | Builder |
get_track_record | How often each signal was right | Free |
get_signal_ledger | Historical signals with outcomes, for backtesting | Builder |
get_recent_events | What changed since your last check | Builder |
get_us_market_session | Can US stocks be traded right now, and which are halted | Free |
get_cross_asset | Are crypto and US equities risk-on together or diverging | Builder |
get_equity_flow | US stock flow from the IEX sample | Builder |
get_service_status | Freshness of every data source | Free |
Readings are context with a measured track record, not trade instructions. Many signals score close to a coin flip; the reliability field says so.
Local stdio server (PyPI)
pip install horus-flow-mcp and set RAPIDAPI_KEY to use the RapidAPI listing from a local MCP client.
Webhook signatures
import hmac, hashlib
def verify(secret: str, timestamp: str, body: bytes, signature: str) -> bool:
expected = "sha256=" + hmac.new(secret.encode(), timestamp.encode() + b"." + body, hashlib.sha256).hexdigest()
return hmac.compare_digest(expected, signature)
# timestamp = request.headers["X-Horus-Timestamp"], signature = request.headers["X-Horus-Signature"]
⚡ 60-Second Quant Bot Quickstart (Python)
import requests
HEADERS = {"X-API-Key": "your_horus_api_key"}
BASE_URL = "https://flow.horustek.pro"
# 1. Query Horus Cortex for the Sovereign Market State
res = requests.get(f"{BASE_URL}/v1/intelligence/cortex", headers=HEADERS).json()
trust_score = res["trust_score"]
policy = res["action_policy"]
boundaries = res["execution_boundaries"]
print(f"🧠 Market State: {res['regime_state']} | Trust: {trust_score}/100")
print(f"🛡️ Action Directive: {policy['directive']}")
print(f"🎯 Support Invalidation: ${boundaries['invalidation_support']:,.0f} | Breakout Target: ${boundaries['breakout_resistance']:,.0f}")
# 2. Enforce Autonomous Risk Rules
if not policy["trend_allowed"]:
print("⛔ Trade Rejected: Cortex forbids trend breakouts during transitions.")
else:
position_multiplier = policy["trend_multiplier"]
# execute_trade(multiplier=position_multiplier, stop_loss=boundaries["invalidation_support"])
💎 Plans
| Feature | Free | Builder / Trader ($29/mo) | Pro ($149/mo) | Institutional ($499/mo) |
|---|---|---|---|---|
| Crypto order flow | Any Binance spot pair | Any Binance spot pair | Any Binance spot pair | Any Binance spot pair |
| Flow history | – | 60 minutes | 24 hours | 24 hours |
| Track record and US session context | ✅ | ✅ | ✅ | ✅ |
| Climate, ignitions, cross-asset, US equities (IEX sample) | – | ✅ | ✅ | ✅ |
| Events (polling and SSE), signal ledger | – | ✅ | ✅ | ✅ |
| Derivatives (10 majors, Binance + Bybit + OKX), composite, Cortex | – | – | ✅ | ✅ |
| Signed webhooks | – | – | ✅ | ✅ |
| Rate limit | 30/min | 120/min | 300/min | 600/min |
| Remote MCP server | ✅ | ✅ | ✅ | ✅ |
Live status and uptime: flow.horustek.pro/status.html
🌐 Live Real-Time Dashboard
Explore the live institutional orderflow visualizer at flow.horustek.pro/dash/.
Engineered with mathematical rigor for Autonomous AI Agents and Quantitative Traders.
© 2026 Horus Tech Ltd. 🦅
mcp-name: pro.horustek/horus-flow-mcp
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.