ConteXtube MCP Server

Connect Claude and other agents to ConteXtube's layered context library through the Model Context Protocol.

What is MCP?

The Model Context Protocol (MCP) lets AI assistants like Claude connect to external data sources. With ConteXtube's MCP server, an assistant can search, retrieve, and assemble structured context, getting the right depth of understanding in minimal tokens.

Quick Start

1. Claude

Add a custom connector with this URL, or add it to claude_desktop_config.json:

{
  "mcpServers": {
    "contextube": {
      "url": "https://contextube.ai/api/mcp"
    }
  }
}

Config location: %APPDATA%\Claude\claude_desktop_config.json (Windows) or ~/Library/Application Support/Claude/claude_desktop_config.json (macOS)

2. Any MCP client

Point any MCP-compatible client to: https://contextube.ai/api/mcp

Streamable HTTP. No key needed. Clients that only speak stdio can use npx mcp-remote https://contextube.ai/api/mcp.

Available Tools

search_content

Search the context library by topic, keyword, or concept. Start here.

// Find content about AI alignment
search_content({ query: "alignment as partnership" })
get_content

Get metadata, layer statistics and the essence of one work.

get_content({ contentId: "peace-treaty-architecture" })
get_layer

Retrieve one depth layer (L1–L5) of a work. Read L1 first.

// L1 = abstract associations, L5 = full narrative
get_layer({ contentId: "peace-treaty-architecture", layer: 1 })
assemble_contextKey feature

Token-budgeted, focus-aware context assembly. Ask for exactly the tokens you need on a specific topic.

// 2000 tokens focused on privacy
assemble_context({
  contentId: "peace-treaty-architecture",
  maxTokens: 2000,
  focus: ["privacy", "surveillance"]
})
list_content

Browse all available works, with filters.

list_flows

List creators' collections.

The 5-Layer Architecture

Each work is decomposed into layers of increasing depth:

L1
Abstract Associations
Gut feelings and associations, written as a symbolic extract
~50 tokens
L2
Ideas & Evidence
Ideas backed by a direct quote from the source
~200 tokens
L3
Connections
Cross-references to other works and domains
~300 tokens
L4
Synthesis
Commentary and analysis: the "so what"
~500 tokens
L5
Full Narrative & Meta
The full source with notes on how the context was made
~1000+ tokens

← Quick understanding · · · Deep knowledge →

Why Not Just Browse the Web?

ConteXtube MCP

  • ✓ Always accessible
  • ✓ 5 depth layers
  • ✓ Token budgets you set
  • ✓ Human-curated meaning
  • ✓ Creator attribution

Web Browsing

  • ✗ Often blocked or fails
  • ✗ Flat text only
  • ✗ Whole pages of raw text
  • ✗ No curation
  • ✗ Attribution lost

RAG / Embeddings

  • ~ If indexed
  • ✗ Flat chunks
  • ~ Varies
  • ✗ Auto-split
  • ✗ Often lost

Example Conversation

You:

“What does Peace Treaty Architecture say about privacy between humans and AI?”

Claude (using ConteXtube MCP):

Searches ConteXtube → finds “Peace Treaty Architecture” → assembles 2,000 tokens focused on privacy → answers with the essay's own reasoning and quotes, credited to its author.

Ready to connect?

Add the MCP server and give your assistant structured understanding.

https://contextube.ai/api/mcp
REST APIhello@contextube.ai