🌐 intelligence-hub-mcp

Multi-Domain Intelligence Intake, 80:20 Crossover Synthesis & Publishing MCP Server

An intelligence hub MCP server that collects the latest trends from multiple cutting-edge fields (AI/vibe coding, synthetic biology, neuroscience, cryptocurrencies, medicine, reverse engineering) in parallel, performs 80:20 Pareto separation and cross-domain crossover analysis, and outputs structured insights.

🏛️ Hermes Ecosystem & Tri-MCP Architecture

This MCP server operates under Hermes AI Agent (Master Orchestrator) as an equal peer alongside academic deep-dive (ocr-mcp-server) and owned-media distribution (news-site-coral).

                    ┌───────────────────────────────┐
                    │      🪐 Hermes AI Agent       │
                    │      (Master Orchestrator)    │
                    └───────┬───────────────┬───────┘
                            │               │
            ┌───────────────┴────┐     ┌────┴──────────────┐
            │ intelligence-hub-mcp│     │  ocr-mcp-server   │
            │  (80:20 Crossover) │     │ (Deep MAD Review) │
            └────────────────────┘     └───────────────────┘
                            │                   │
                            └─────────┬─────────┘
                                      ▼
                    ┌───────────────────────────────┐
                    │       news-site-coral         │
                    │  (WIRED Narrative Publishing) │
                    └───────────────────────────────┘

1. Multi-layer Intake (intake/):

   • Type A (RSS/Atom): bioRxiv (SynthBio / Neuroscience), medRxiv, Zenn, Ethereum Research, Paradigm

   • Type B (Public API): Hacker News (Firebase API), arXiv (Atom API)

   • Type C (Browser/undocumented API): GitHub Trending (Phase 2)

   • Smart skip: automatically prevents unnecessary frequent requests based on interval_hours judgment

2. Intellectualization & Persistence (storage/ & analysis/):

   • LanceDB (PyArrow): persists IntelligenceRecord metadata + 768-dim vector embeddings

   • 80:20 Pareto Analysis: 80% fixed-point observation + 20% serendipity enclave

   • Crossover extraction: uses LLM to extract structural analogies and mutual applicability across different genres

3. Publish & Integration (mcp/):

   • Attaches suggested_themes / suggested_tags to structured data (CrossoverDigest) and delivers via Hermes Agent to X (140-character thread) / Note (essay) / Coral Magazine (WIRED-style 3-tier HTML) in a multi-channel manner.

Related MCP server: worldmonitor-mcp

🚀 Quick Start

Dependency Synchronization

uv sync --all-groups

Running Tests

uv run pytest

Layer Boundary Verification (Spaghetti Guard)

npx @naoya.k/spaghetti-guard check

Running the Periodic Collection CLI (Cron / Batch Execution)

uv run intelligence-hub-cron

Starting the MCP Server (Stdio Transport)

uv run intelligence-hub-mcp

🛠️ Public MCP Tools

• fetch_intelligence_feed(category=None, limit_per_channel=None, force_fetch=False): Fetch the latest feeds in parallel and save to LanceDB (with interval_hours judgment)

• search_intelligence(query, category=None, limit=10): Semantic vector similarity search

• list_intelligence_records(category=None, is_serendipity=None, limit=50): Filtered list of saved articles

• generate_crossover_digest(period="daily", fetch_latest_first=True): Generate cross-domain crossover report & structured data with Coral recommended tags

• get_feed_status(): Check monitored feed configuration, interval status, and LanceDB statistics