diff --git a/modules/ROOT/pages/use-mulesoft-docs-with-ai.adoc b/modules/ROOT/pages/use-mulesoft-docs-with-ai.adoc index 27db54c77..8d4d3ffd9 100644 --- a/modules/ROOT/pages/use-mulesoft-docs-with-ai.adoc +++ b/modules/ROOT/pages/use-mulesoft-docs-with-ai.adoc @@ -6,27 +6,61 @@ endif::[] MuleSoft documentation is optimized for AI. Whether you're a human pasting context into an LLM or an AI agent ingesting documentation autonomously, our content is structured to save tokens, stay grounded, and maximize technical accuracy. +MuleSoft documentation offers several ways to feed content to AI depending on your workflow. + +[cols="1,2,2", options="header"] +|=== +| Approach | Best for | When to use + +| <> +| Grabbing a single page by hand +| You're reading a specific page and want to copy or send it to an LLM quickly, with no setup. + +| <> +| Live, interactive lookups in an AI assistant +| You want your AI assistant to access official documentation for you as you work. Your MCP-compatible client (such as Claude, Cursor, or VS Code) queries the documentation directly, with no files or infrastructure to manage. + +| <> +| Bulk ingest into your own pipeline, or browsing the documentation structure +| You're building a RAG pipeline or vector database and want to download and index the documentation on your own infrastructure, or you want a full, browsable index of a product's documentation. +|=== + [[page-options]] == Page Options Every page of MuleSoft documentation includes a dropdown menu for AI consumption. -image::ai-context-menu.png["Dropdown menu with options Copy page URL, Copy page content, and Download page content for AI use"] +image::ai-context-menu.png["Dropdown menu with options Copy as Markdown, View as Markdown, View on GitHub, View llms.txt, Use Docs MCP Server, and Open in ChatGPT, Claude, or Perplexity"] // CHANGE IMAGE Use these options to quickly grab clean text or jump straight into your workflow: * *Copy as Markdown*: Copies the Markdown version of the page to your clipboard to streamline your prompt engineering. * *View as Markdown*: Opens the Markdown version of the page in a new tab for sharing or inspecting. * *View on GitHub*: Opens the page's source on GitHub to review, suggest changes, or log issues. +* *Use Docs MCP Server*: Connects your AI assistant to MuleSoft documentation through the Salesforce Docs MCP Server. See <>. * *View llms.txt*: Opens the sitewide `llms.txt` index file. See <>. * *Open in ChatGPT / Claude / Perplexity*: Opens a conversation with the current page as context so the LLM can answer your questions instantly. +[[docs-mcp-server]] +== Docs MCP Server + +The Salesforce Docs MCP Server gives AI assistants live access to MuleSoft documentation through the https://modelcontextprotocol.io/[Model Context Protocol (MCP)]. Instead of relying on outdated training data or unstructured web scrapes, your MCP-compatible client can discover, search, fetch, and ask questions about MuleSoft documentation directly. It requires no API key and no sign-up, and it refreshes daily. + +Connect any MCP-compatible client to: + +[source,text] +---- +https://mcp.docs.salesforce.com/ +---- + +For setup instructions, the full tool reference, and example prompts, see https://mcp.docs.salesforce.com/[Salesforce Docs MCP Server]. + [[llms-txt]] -== llms.txt and llms-full.txt +== llms.txt and llms-full.txt MuleSoft documentation publishes `llms.txt` and `llms-full.txt` index files, which are machine-readable Markdown catalogs that let AI tools and RAG pipelines cleanly discover, search, or bulk-ingest our documentation without crawling raw HTML. -For example, you can point an AI coding assistant to the sitewide or product `llms.txt` URL for real-time integration context, or import the `llms-full.txt` file into an vector database to ground AI agents. +For example, you can point an AI coding assistant to the sitewide or product `llms.txt` URL for real-time integration context, or import the `llms-full.txt` file into a vector database to ground AI agents. Depending on your data scope, you can target three types of endpoints: