Ask about anything in HelpGuides.io Documentation

Type a question in your own words below. Answers are written from our published articles and cite the ones they came from.

Recently updated

    ↑↓ select · ↵ open · esc close
    Dashboard
    Edit Article Logout

    Model Context Protocol

    Important

    You may have been redirected to this page when attempting to access the /mcp path to your HelpGuides.io documentation.

    What is Model Context Protocol

    Model Context Protocol (MCP) is an emerging open standard that defines how AI systems interact with external tools and knowledge sources in a structured, predictable, and programmatic way.

    Rather than relying on scraping web pages or processing unstructured text, MCP allows AI clients to discover available capabilities, query systems using well-defined inputs, and retrieve results in consistent, machine-readable formats.

    This enables AI assistants, agents, and retrieval-augmented generation (RAG) workflows to integrate external context reliably, turning documentation and services into first-class components of an enterprise AI stack while reducing fragility, ambiguity, and long-term integration risk.

    MCP Support Simplifies Integration for RAG-based scenarios

    Content published in HelpGuides.io is automatically structured and chunked at publish time. Long articles are broken into logically related sections based on headings, then tokenized and indexed as discrete units.

    This approach allows search to retrieve only the most relevant portions of content for a given query, improving response accuracy.

    You can see this when searching for content in the UX as search can show relative sections. For example, there are multiple results returned for the article Setting up Text Message Marketing:

    Search results showing several sections of the same article, each with its own heading and excerpt

    With support for Model Context Protocol (MCP), downstream AI tools, such as Microsoft Copilot when integrated via MCP, benefit from this same structure content approach. And most importantly, no longer need to rely on scraping HTML.

    Instead, documentation is exposed through a clean, structured, and pre-chunked interface, ensuring that only the content most relevant to a given query is returned. In practice, MCP enables applications to interact with structured knowledge sources more intelligently, retrieving precise context optimized for retrieval-augmented generation (RAG).

    Turning on MCP and finding your URL

    Open your project, click AI Settings and stay on the Model Context Protocol tab. Make sure Enable Model Context Protocol is on, then copy Model Context Protocol URL for this project. It's your site's domain followed by /mcp:

    https://YOUR-DOMAIN/mcp

    AI assistants sign in with OAuth: the first time you connect one, you sign in to HelpGuides and approve its access. When the switch is off, the URL stops responding and connected assistants lose access. See AI Settings and, for every tool the endpoint offers, Model Context Protocol Endpoints.

    AI Settings Model Context Protocol tab with the MCP URL and the Enable switch
    AI Settings, Model Context Protocol tab

    Connecting your MCP to LLMs

    You can connect your MCP URL to any AI assistant that supports remote MCP servers. See the instructions for:

    Testing MCP via CURL

    You can test your MCP endpoint with curl. Every request needs an OAuth access token in an Authorization: Bearer header. Without one, the endpoint replies 401 with a WWW-Authenticate header that tells MCP clients where to sign in.

    Note, in practice, these interactions are handled automatically: an MCP client queries the server, discovers the supported capabilities and endpoints, and then invokes the appropriate endpoint to fulfill a request.

    Initial request

    The first request initializes a session. It returns the server info and an Mcp-Session-Id header, which you send with every later request:

    curl -i -X POST https://YOUR-DOMAIN/mcp \ -H "Authorization: Bearer ACCESS_TOKEN" \ -H "Content-Type: application/json" \ -H "Accept: application/json, text/event-stream" \ -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"Example","version":"1.0.0"}}}'

    The response includes the Mcp-Session-Id header:

    HTTP/1.1 200 OK Content-Type: text/event-stream Date: Fri, 16 Jan 2026 17:25:08 GMT Server: Kestrel Cache-Control: no-cache,no-store Content-Encoding: identity Transfer-Encoding: chunked Content-Security-Policy: frame-ancestors 'none' X-Frame-Options: DENY Mcp-Session-Id: MM2ZUNNpVih7r0TGZb38pg event: message data: {"result":{"protocolVersion":"2024-11-05","capabilities":{"logging":{},"tools":{"listChanged":true}},"serverInfo":{"name":"io.helpguides/sever","version":"1.0.0"}},"id":1,"jsonrpc":"2.0"}

    Perform a query

    Next, call a tool using the session ID. This example calls the search tool:

    curl -X POST https://YOUR-DOMAIN/mcp \ -H "Authorization: Bearer ACCESS_TOKEN" \ -H "Content-Type: application/json" \ -H "Accept: application/json, text/event-stream" \ -H "Mcp-Session-Id: YOUR_MCP_SESSION_ID" \ -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"search","arguments":{"s":"setting up SMS marketing"}}}'

    For the full list of tools and their parameters, see Model Context Protocol Endpoints. To have an agent act on instructions you leave in the editor, see Review Comments and AI Agents.

    How helpful was this article?

    👍 or 👎

    Related Articles

    Markdown Version