> ## Documentation Index > Fetch the complete documentation index at: https://docs.flowx.ai/llms.txt > Use this file to discover all available pages before exploring further. # Using FlowX docs with AI assistants > Connect this documentation to Claude, ChatGPT, Cursor, or any MCP client so your assistant answers from the current docs instead of guessing. This documentation is built to be read by AI assistants as well as people. You can connect it to your assistant so that answers come from the current FlowX.AI documentation rather than from whatever the model happened to memorize during training. > [!NOTE] > > This page is about connecting **the documentation site** to an AI assistant. It is not about the MCP Server data source you configure inside a project. For that, see [MCP integration overview](https://docs.flowx.ai/6.x/docs/platform-deep-dive/integrations/mcp-integration/mcp-integration-overview.md). *** ## Why connect the docs Language models are trained at a point in time and FlowX.AI ships monthly. An assistant answering from memory will confidently describe features that changed, node types that were renamed, and configuration keys that never existed. Connecting the docs changes where the answer comes from. The assistant searches the live documentation, reads the actual page, and cites it back to you. > [!TIP] > > This matters most for version-sensitive questions. Availability, environment variables, and node configuration differ between releases, and those are exactly the details a model is most likely to get wrong from memory. *** ## Connect from any page Every page on this site has a contextual menu at the top with the connection options built in. **1. Open the menu on any documentation page** Use the control at the top of the page, beside the page title. **2. Pick your assistant or editor** Options include Claude, ChatGPT, Perplexity, Cursor, and VS Code. Each one hands the current page, or the connection details, straight to that tool. **3. Ask your question** The assistant now has the documentation available and will search it rather than answering from memory. The same menu also offers **Copy page** and **View as Markdown** if you would rather paste content in manually. *** ## Connect manually over MCP For any client that supports the Model Context Protocol, connect to this endpoint over HTTP: ``` https://docs.flowx.ai/mcp ``` In Claude Code, add it with: ```bash theme={"dark"} claude mcp add --transport http flowx-docs https://docs.flowx.ai/mcp ``` Other clients store MCP server definitions in their own configuration file, and most accept a remote HTTP server as a name and URL pair. Use the contextual menu on any page to get the exact configuration for Cursor and VS Code, since it generates the correct file and format for you. ### What the assistant can do once connected The docs server exposes three tools. | Tool | What it does | | - | - | | `search_flow_x_ai` | Searches the documentation and returns matching content with direct links to the source page | | `query_docs_filesystem_flow_x_ai` | Reads and searches the documentation as a read-only file tree, for exact keyword matching or reading a full page | | `submit_feedback` | Reports a documentation problem to the docs team, tied to the page path it concerns | > [!NOTE] > > `query_docs_filesystem_flow_x_ai` runs against a sandboxed, in-memory copy of the documentation. It is not a shell, and nothing runs on your machine, the docs server, or any FlowX environment. > [!TIP] > > `submit_feedback` is worth knowing about. If your assistant gives you an answer that turns out to be wrong because a page is outdated or incomplete, you can ask it to report that, and the correction reaches the docs team with the page path attached. *** ## Install the FlowX skill The docs also publish a skill: a short capability summary that tells an agent what it can build with FlowX.AI, which terms to use, and which constraints apply. Search answers one question at a time. The skill gives the agent working context before you ask. If you connected over MCP, you already have it. The docs server exposes the skill as an MCP resource, so your client can load it without a separate install. For an agent that does not use MCP, install it with the skills CLI: ```bash theme={"dark"} npx skills add https://docs.flowx.ai ``` The CLI requires a recent Node.js release (22.20 or later). The raw file is published at `https://docs.flowx.ai/skill.md`. > [!NOTE] > > The skill describes the current LTS release family. On any other release, including 6.0, rely on search results from your version's documentation tree rather than the skill's capability summary. ### Set it up with a prompt Copy this prompt into your agent, or open it in Cursor, and let the agent do the setup. Connect the FlowX.AI documentation to your agent. Connect the FlowX.AI documentation to this agent so you answer FlowX.AI questions from the current docs instead of from memory. 1. Add the remote MCP server `https://docs.flowx.ai/mcp` over HTTP, using the configuration format this client expects. 2. If this client does not load MCP resources, install the FlowX.AI skill with `npx skills add https://docs.flowx.ai`. 3. Ask me which FlowX.AI release I am running, and search that version's documentation tree in every answer. 4. Confirm the setup by searching the docs for "build your first process" and returning the page link. *** ## Bulk formats If you are building a tool, seeding a knowledge base, or feeding a long-context model, three formats are published. > [card] Page as Markdown > > Append `.md` to any documentation URL to get that single page as clean Markdown. > [card] llms.txt > > An index of the documentation with a one-line description per page. Good for routing and discovery. > [card] llms-full.txt > > The complete documentation as a single Markdown file. Large, and intended for ingestion rather than browsing. > [card] Sitemap > > `https://docs.flowx.ai/sitemap.xml` lists every indexed page URL. Example, fetching one page as Markdown: ```bash theme={"dark"} curl https://docs.flowx.ai/6.x/docs/getting-started/glossary.md ``` *** ## Working across versions FlowX.AI documentation is published as one tree per release family, and the version is part of every URL. When you ask a version-specific question, name the release in your prompt so the assistant searches and cites the right tree. > [!WARNING] > > Ask for the version you are running, and check the version in the links your assistant returns. A question asked without a version can be answered from any tree, including release families that have reached end of life. Feature Release content documented in the current LTS tree carries a **SaaS** availability badge. Those features are live on managed deployments immediately, and reach self-hosted deployments with the next LTS release family. If your assistant reports a feature as available, confirm against that badge before planning a self-hosted rollout. *** ## Related resources > [card] [MCP integration overview](https://docs.flowx.ai/6.x/docs/platform-deep-dive/integrations/mcp-integration/mcp-integration-overview.md) > > Configure an MCP Server as a data source inside a FlowX project > [card] [Glossary](https://docs.flowx.ai/6.x/docs/getting-started/glossary.md) > > Core FlowX.AI terms, so your prompts use the right vocabulary > [card] [Build your first process](https://docs.flowx.ai/6.x/docs/getting-started/building-your-first-proc.md) > > Start here if you are new to the platform > [card] [Knowledge Base and RAG](https://docs.flowx.ai/6.x/ai-platform/patterns/knowledge-base-rag.md) > > Build retrieval into your own FlowX agents ## Related topics - [FlowX.AI data architecture: which service uses which store](https://docs.flowx.ai/6.x/docs/platform-overview/data-architecture.md) - [Using AI agents](https://docs.flowx.ai/6.x/ai-platform/using-agents/overview.md) - [FlowX.AI architecture: services, layers, and data flow](https://docs.flowx.ai/6.x/docs/platform-overview/flowx-architecture.md) - [AI Gateway setup](https://docs.flowx.ai/6.x/setup-guides/ai-gateway-setup.md) - [FlowX.AI 5.3.0 Release Notes](https://docs.flowx.ai/release-notes/v5.x/v5.3.0-december-2025/v5.3.0-december-2025.md) This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.