Model Context Protocol

Connect AI Assistants to Your Documentation

Model Context Protocol (MCP) enables Claude, ChatGPT, Cursor, and other AI assistants to access your Contextium documentation directly. No more copy-pasting. No more context switching.

One command setup
npx @contextium/mcp-server@latest setup
UNDERSTANDING MCP

What is Model Context Protocol?

MCP is an open standard that lets AI assistants securely access external data sources. Think of it as OAuth for AI - secure, standardized, and simple.

Without MCP

  • Copy documentation from your knowledge base
  • Paste into AI assistant chat
  • Hope you included all relevant context
  • Repeat for every question

⏱️ 5 minutes wasted per question

With MCP

  • Ask AI assistant your question directly
  • AI automatically searches your Contextium docs
  • Get accurate answer with citations
  • Works every time, automatically

⚡ Instant answers from your docs

How Contextium MCP Works

Three simple steps to connect your documentation to any AI assistant

1

Install MCP Server

Run our one-line setup command. It automatically configures your AI assistant to connect to Contextium via secure OAuth.

2

Authorize Access

Browser opens automatically for secure authorization. Grant your AI assistant read-only access to your workspace documentation.

3

Start Asking

Your AI assistant can now search and reference your documentation automatically. Ask questions and get answers based on YOUR docs.

Setup Guide

Choose your AI assistant and follow the setup instructions

Claude Desktop Setup

1Run Setup Command

npx @contextium/mcp-server@latest setup

This command automatically configures Claude Desktop and opens your browser for authorization.

2Authorize in Browser

Your browser will open automatically. Log in to Contextium and authorize Claude to access your workspace documentation. This is a one-time setup with secure OAuth.

3Restart Claude Desktop

Close and reopen Claude Desktop. You'll see a 🔧 icon indicating MCP tools are available. Try asking: "Search my Contextium documentation for authentication guides"

Pro Tip

Claude can now automatically search your docs when you ask questions. No need to explicitly tell it to search Contextium!

What You Can Do With MCP

Real-world use cases for developers and teams

Instant Documentation Search

Ask Claude "How does our authentication work?" and get answers pulled directly from your internal docs.

Code Generation from Specs

"Generate a React component based on our design system docs" - AI reads your guidelines and generates compliant code.

Onboarding Assistance

New team members can ask questions about your codebase, architecture, and processes without interrupting seniors.

API Integration Help

"Show me how to authenticate with our API" - get step-by-step guidance with actual code examples from your docs.

Code Review Context

During code review, ask AI to check if implementations match your documented standards and best practices.

Documentation QA

Ask "Is this feature documented?" or "Find all references to OAuth" - AI searches your entire knowledge base instantly.

TECHNICAL DETAILS

How MCP Works Under the Hood

Architecture

MCP uses a client-server architecture where your AI assistant (client) communicates with Contextium's MCP server through a standardized protocol. The server provides secure access to your documentation via defined tools and resources.

  • Protocol: JSON-RPC 2.0 over stdio or HTTP
  • Authentication: OAuth 2.0 with PKCE flow
  • Permissions: Read-only access, scoped to workspace

Available MCP Tools

Contextium's MCP server exposes these tools that AI assistants can use:

Discovery & Search

list_workspaces

List all accessible workspaces

list_projects

List all context libraries in workspace

list_agents

List all agents in workspace

list_skills

List all skills in workspace

list_files

List files in a context library or agent

list_tags

List all tags in workspace

search_documentation

Full-text search across all workspace documents

search_by_tags

Find files by tag combinations

Read & Retrieve

get_file

Retrieve complete file content by ID

get_file_versions

Get version history for a file

get_project_structure

Get folder structure of a context library

Create Resources

create_project

Create a new context library in workspace

create_agent

Create a new agent with custom instructions

create_skill

Create a new reusable skill

create_file

Create a new documentation file

Update & Delete

update_file

Update file content with version tracking

delete_file

Delete a file from a context library

Workflows

list_workflows

List all workflows in workspace

load_workflow

Load all resources from a workflow (agents, libraries, files, tags)

Security & Privacy

  • Read-only access: AI assistants can only read documentation, never write or modify
  • Workspace isolation: Each authorization is scoped to a single workspace
  • Revokable access: Revoke access anytime from Contextium settings
  • Encrypted communication: All data transmitted over HTTPS
  • No data retention: AI assistants don't store your documentation

Frequently Asked Questions

What AI assistants support MCP?

Currently, Claude Desktop, Cursor IDE, and several other AI assistants support MCP. ChatGPT support is coming soon. Any AI assistant that implements the Model Context Protocol specification can connect to Contextium.

Is my documentation secure with MCP?

Yes. MCP connections are read-only, workspace-scoped, and use OAuth 2.0 for authentication. Your documentation never leaves Contextium's servers - AI assistants only receive search results and specific documents they request. You can revoke access anytime.

Does MCP work with private documentation?

Absolutely. MCP is designed for private, internal documentation. Only authenticated users with workspace access can authorize AI assistants, and those assistants can only access documentation within that specific workspace.

How much does MCP cost?

MCP integration is included in all Contextium plans, including the free Starter plan. There are no additional charges for MCP connections or API calls.

Can I build custom MCP tools?

Yes! Contextium's MCP server is open source. You can extend it with custom tools, or build your own MCP server that integrates with Contextium's API. Check our documentation for the MCP SDK and examples.

What happens if I disconnect MCP?

You can disconnect MCP access anytime from Contextium Settings → Integrations. Once disconnected, AI assistants will no longer be able to search or access your documentation. You can reconnect at any time by running the setup command again.

Ready to Connect Your Docs to AI?

Start using MCP with Contextium today. Free plan includes full MCP support.

No credit card required • Setup in under 2 minutes