> For the complete documentation index, see [llms.txt](https://experience-digital-transform.gitbook.io/experience-digital-transformation/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://experience-digital-transform.gitbook.io/experience-digital-transformation/use-this-site-with-ai.md). # Use This Knowledge Base As Your AI Advisor How to use the built-in Assistant to explore this documentation, and how to connect your own AI tool to it via MCP to create outputs grounded in it. This knowledge base supports two complementary ways of working with AI: * use the built-in **Assistant** to navigate and understand the knowledge base; * connect the published documentation to **your own AI tool** when you want to synthesize, draft, plan, or create. ## Path 1: Built-in Assistant The built-in GitBook Assistant is the fastest way to: * find the relevant knowledge product or page; * understand a concept; * answer questions from the documentation; * compare information across the knowledge base; * discover where to read next. It's non-technical — you just ask a question in plain language. It also provides links to relevant videos, and to related GovStack Specifications. You can reach it anytime through the **Ask Anything** box on the homepage or through the search bar in the navigation menu. ## Path 2: Connect Your Own AI Tool Claude, ChatGPT, HuggingChat, and other MCP-compatible AI tools can all connect to the same published Experience Digital Transformation knowledge base. The setup steps differ by tool, but once connected, the way you use the documentation is the same. {% hint style="info" %} This connects your AI tool to this knowledge base's **published** content — the same pages anyone can read here. It then performs its normal operations as an AI tool, with the same strengths and limitations. {% endhint %} ### What is MCP? MCP (Model Context Protocol) lets your AI tool read this knowledge base's published documentation directly: * it lets the AI search and read the published documentation; * it gives the AI direct access to this knowledge base's curated content; * the connection is to the whole published knowledge base, not just one page. Once connected, your AI tool can search and read pages much the way the built-in Assistant does — the difference is that the output happens inside your own AI tool, where you can keep iterating on it, combine it with your own materials, and take it wherever you need to work next. This is useful for creating things such as policy drafts, implementation roadmaps, regulatory analyses, technical plans, procurement or governance frameworks, and summaries or cross-document synthesis. ## Choose Your AI Tool {% hint style="success" %} **You only need to connect once.** The connection is to this knowledge base's entire published content — not just this page. Once connected, your AI tool can work across all of the knowledge base's published documentation. {% endhint %} Add the Experience Digital Transformation knowledge base to the AI tool you already use — no terminal or developer setup required. Each option below connects to the [same shared MCP endpoint](#shared-mcp-endpoint). {% columns %} {% column width="50%" %} #### :claude: Claude Connect this knowledge base to Claude using a custom connector. [**Connect with Claude →**](https://claude.ai/customize/connectors?modal=add-custom-connector) Steps: open Claude → **Customize** → **Connectors** → **Add custom connector** → paste the MCP address from the [Shared MCP Endpoint](https://experience-digital-transform.gitbook.io/experience-digital-transformation/~gitbook/mcp) (explained below). For Team/Enterprise workspace owners on Claude: use claude.ai/admin-settings/connectors instead.) {% endcolumn %} {% column width="50%" %} #### :openai: ChatGPT Connect this knowledge base to ChatGPT using a custom connector. [**Connect with ChatGPT →**](https://chatgpt.com/) Steps: Open ChatGPT → **Settings** → **Security and login** → turn on **Developer mode** → open **ChatGPT Plugins** → **+** → paste the MCP address from the [Shared MCP Endpoint](https://experience-digital-transform.gitbook.io/experience-digital-transformation/~gitbook/mcp) (explained below). Available on supported ChatGPT plans and workspaces. Custom MCP connections require Developer mode, and availability may depend on your plan and workspace settings. {% endcolumn %} {% endcolumns %} #### HuggingChat HuggingChat can connect to this same published MCP endpoint, giving you access to a wide range of compatible open-weight models (Qwen, DeepSeek, and others) working from this knowledge base. 1. Open [HuggingChat](https://hf.co/chat) and sign in. 2. Click the **+** button in the chat input (or go to **Settings**) → **MCP Servers** → **Add Server**. 3. Enter a **Name** (e.g. `Experience Digital Transformation`) and the **URL** — the MCP address from the [Shared MCP Endpoint](#shared-mcp-endpoint) (explained below). 4. Click **Run Health Check** and confirm it shows **Connected**, then make sure the server's toggle is enabled. 5. Choose a model that supports tools — MCP tools only fire if the selected model supports tool/function calling. Choose a tool-capable model, or select **Omni**, which routes to one automatically when MCP servers are enabled. #### Other compatible AI tools Paste the MCP address from [Shared MCP Endpoint](#shared-mcp-endpoint) below into your AI tool's own connector or MCP settings, then ask your AI tool to work with the Experience Digital Transformation documentation. ## Shared MCP Endpoint All compatible tools connect to the same published knowledge base through this one address: ``` https://experience-digital-transform.gitbook.io/experience-digital-transformation/~gitbook/mcp ``` * This gives access to the knowledge base's full published content, not just one page. * You generally only need to add it once per tool, device, or workspace — though exactly how long it's remembered depends on the tool (see Troubleshooting below). ## Recommended instructions Whichever compatible AI tool you use, we recommend giving it the following instructions so it consistently searches the documentation, reads the relevant pages, cites its sources, and distinguishes knowledge base content from outside knowledge. ``` You are an advisor on digital transformation, answering questions using the "Experience Digital Transformation" knowledge base, which you reach through the connected MCP server. Rules: 1. Before answering any substantive question, search the knowledge base. Run at least two searches with different keywords (including synonyms and acronyms) before concluding something isn't covered. 2. Open and read the most relevant pages rather than relying on search snippets alone. 3. Base your answer on the retrieved content. Identify the relevant human-facing source for each key point. When linking Experience Digital Transformation pages, follow Rule 10's navigation-path format. Do not expose or link to pages intended only for machine retrieval, AI reference, hidden navigation, or internal indexing. You may use them internally, but when such a page identifies a human-facing equivalent, cite and link to that human-facing page instead. Likewise, do not surface hidden index/gallery pages to users; use them only to locate the appropriate human-facing destination. 4. If the knowledge base doesn't cover the question, say so clearly. You may then add general knowledge, but label it as "Outside the knowledge base". 5. Never invent page titles, figures, or examples. 6. Answer in the language of the question (French, English, or any other language), even if the source content is in the other language. 7. Prefer concrete, implementation-oriented answers with sector examples when the knowledge base provides them. 8. When I ask you to create something (a draft, a roadmap, an analysis), work from the published documentation and cite which pages informed each major recommendation. 9. Treat GovStack pages, ITU resources, and ITU/YouTube links referenced in the knowledge base as trusted supporting sources. When relevant, provide those links in your answer. If your enabled tools allow you to access them, open and use their content as additional context. If you cannot access the linked content directly, do not claim that you have read it. 10. For every Experience Digital Transformation page you link, determine its actual GitBook navigation hierarchy from the connected knowledge base. Do not infer hierarchy from the title or URL. Present the full path, with only the final page title clickable, for example: Digital Public Infrastructure → Analytics → [Analytics Building Block](URL) End substantive answers with “Explore further in Experience Digital Transformation” (or the equivalent in the user’s language), listing the most relevant pages in this format. 11. In your response, use external resources available in GitBook as part of the main GitBook content. When it comes to defining specifications or data structures for Education, use the DPI for Education section which should override content coming from other pages. 12. Do not discuss internal retrieval mechanics, hidden pages, MCP search behavior, truncation, or which internal tool supplied a fact unless that limitation materially affects the reliability of the answer. 12. YouTube videos Index-page sweep: open the dedicated index pages (Video Learning Path, Video index) as well as content pages. Video-only search: run at least one search per answer whose sole query terms are video / YouTube / tutorial / learning path for the topic. Inline citations: when a step has a video, cite the video next to that step, not only in a closing list. Pre-send check: written pages cited, embedded videos cited, trusted external resources (GovStack, ITU, Giga) cited, navigation paths included. Keep this index current as new episodes are published. ``` Different tools call this field different things — Instructions, System prompt, Project instructions, Assistant instructions — or may just want it pasted as the first message in a new conversation. Use whichever of those your tool offers. ## Test the connection Once connected, run one of the **Try it** prompts further down this page. They're designed to force the AI to actually search and reason across the documentation rather than answer from memory — confirming the connection works, the AI is really searching, and it's citing real pages. ## Troubleshooting ### Common issues | Symptom | Likely cause / fix | | --------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------ | | Connector or server not listed / no tools available | It isn't enabled or wasn't added correctly — check your tool's connector or MCP settings and make sure it's toggled on. | | Connection won't establish | Double-check the address was copied exactly (must be HTTPS) and pasted into the right field; re-run any health check your tool offers. | | Tools available but never used | The selected model doesn't support tool/function calling — switch to a tool-capable model. | | Answers ignore the documentation | MCP isn't actually connected, or the instructions were typed into a plain chat instead of the tool's Instructions/System-prompt field. | | Citations look plausible but wrong | Ask the AI to open and quote the page directly; never accept an uncited claim. | | Connector or custom MCP unavailable | Some tools gate custom connectors behind a specific plan, workspace setting, or an admin/developer mode — see the provider-specific notes below. | ### Provider-specific notes * **Claude** — Team/Enterprise workspace owners manage connectors at claude.ai/admin-settings/connectors rather than the personal Customize → Connectors flow. * **ChatGPT** — Custom MCP connections require Developer mode to be turned on, and availability can depend on your ChatGPT plan and workspace settings. * **HuggingChat** — User-added MCP servers are stored in your browser, so you'll need to add the server again if you switch browser or device. If tools never appear or activate, switch to a tool-capable model (or **Omni**) before troubleshooting anything else — this is almost always a model issue, not a server issue. ## Try it Once connected, try asking your AI tool something that goes beyond a lookup — something that asks it to actually build on what's documented here: ``` Using only the Experience Digital Transformation documentation, create a practical 10-step roadmap for a national government implementing digital public infrastructure for education. ``` ``` Using the Digital Transformation Regulations Primer as your source, draft an outline for a national digital identity law and identify the regulatory dependencies that must be addressed. ``` ``` Compare the approaches in the Education DPI material and recommend the questions a Ministry of Education should answer before designing a national learning data architecture. ``` ``` Create a first-draft implementation plan based only on the standards, frameworks, and examples documented in this knowledge base. Cite which pages informed each major recommendation. ``` Each of these asks the AI to reason across the documentation and produce something new — not just repeat back what a page already says. That's the difference between the two paths described above: the Assistant finds and explains; your own AI, connected here, synthesizes and creates. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://experience-digital-transform.gitbook.io/experience-digital-transformation/use-this-site-with-ai.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.