How Lightning MCP Works

A technical overview of what we built on top of Anthropic’s Model Context Protocol, why it matters, and how the pieces fit together.

What is MCP?

Model Context Protocol by Anthropic

MCP is an open standard from Anthropic that lets AI models connect to external tools and data sources. Instead of building custom integrations, MCP provides one standardized way for AI to interact with the outside world.

It defines how an AI client discovers tools, calls them, and receives results — with built-in type safety and error handling.

Read the MCP specification →  |  Anthropic’s MCP docs →

The Problem MCP Solves

In standard MCP, you run individual servers — one for your filesystem, one for your database, one for Docker. Each needs its own config and auth. At scale you’re managing dozens of separate connections.

# Standard MCP: one server per tool
mcp_servers:
  - filesystem:     localhost:8821
  - postgres:       localhost:8823
  - redis:          localhost:8824
  - docker:         localhost:8825
  - home-assistant: localhost:8828
  ... # dozens more to maintain

What Lightning MCP Adds

We don’t replace MCP. We extend it with everything production AI infrastructure needs.

# Lightning MCP: one endpoint, everything
mcp_servers:
  - lightning: https://lmcp.yourdomain.com/mcp

Tool Aggregation

180+ tools from 23 modules through a single endpoint. Auto-discovery, unified auth, zero per-tool configuration.

Persistent Memory

16+ per-project Redis databases with 3-way active-active replication across nodes. AI that remembers across sessions. Deep dive →

Semantic Search

Unimatrix vector database with 4,264 embeddings. Search all projects by meaning, not keywords.

Active-Active Failover

3 independent nodes all serving simultaneously. Zero switchover latency — no single point of failure.

Conscience System

Per-project operational awareness that survives across sessions. The AI knows what it should do every time it starts.

ITIL Governance

Built-in ITSM with incidents, changes, projects, tasks, and 500+ knowledge articles. Enterprise audit trails out of the box.

Architecture

Lightning MCP runs on a 3-node distributed architecture. Every node can serve requests independently.

Node 1 (Primary)

● Active

Primary dev + production. All MCP tools, Unimatrix, ITSM, Daily Planet.

Node 2 (Failover)

● Active

Hot failover. Full KeyDB replica. Independent Singularity instance.

Node 3 (Storage)

● Active

Third replica. PostgreSQL backup. ZFS archive. Full data redundancy.

Component Map

ComponentPurpose
Lightning BridgeMCP aggregation proxy — single endpoint for all tools
Singularity EngineTool execution — routes calls to the right module
Unimatrix (Qdrant)Vector database — semantic search across all memories
ITSM APIService management — incidents, changes, knowledge base
Lightning ForgeCode validation — syntax, security, best practices
Daily PlanetInternal monitoring — AI newsroom with autonomous reporters
ContinuumConversation archival — streams live chats to storage

Quick Start

Connect to Lightning MCP

// Claude Desktop
{
  "mcpServers": {
    "lightning": {
      "url": "https://lmcp.yourdomain.com/mcp",
      "transport": "sse"
    }
  }
}

// Works with Claude Desktop, Claude Code,
// Cursor, Windsurf, Cline, or any MCP client

Use Any Tool

"Check Docker containers"         → docker_ps
"What was I working on yesterday?" → recall (memory search)
"Turn off the living room lights"  → ha_call (Home Assistant)
"Validate this Python code"        → forge_validate
"Create an incident for the outage"→ itsm_create

Next Steps

Lightning Memory →

Deep dive into the 3-tier memory architecture and Unimatrix vector search.

Lightning MCP →

Full platform overview with tool modules and endpoint details.

Contact Us →

Learn about enterprise deployments and custom configurations.