> For the complete documentation index, see [llms.txt](https://docs.scribeware.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.scribeware.com/into-the-weeds/ai-assistant-mcp.md). # The AI Assistant (MCP) ScribeWare can hand an AI assistant a direct line into your data. Instead of copying text back and forth into a chat window, you point your AI tool at ScribeWare once, and from then on you can just say *"tighten up the observations in the Roof chapter"* — and it reads the report, makes the edits, and they show up in the app. The connection uses **MCP** (Model Context Protocol), an open standard that most modern AI tools now speak. You don't need to know anything about it beyond copying one snippet of setup text, which ScribeWare generates for you. Find it under **Settings → AI**. {% hint style="warning" %} The AI Assistant is in **beta**. It works, and inspectors are using it daily, but expect rough edges — and expect the AI to occasionally get something wrong. Everything it changes can be reverted from the Changes bar, so nothing it does is one-way. {% endhint %} ## What It Can Actually Do The AI isn't looking at your screen or clicking buttons. ScribeWare gives it around fifty specific abilities, and it can only do what's on that list: | Area | What the AI can do | | ----------------------- | ------------------------------------------------------------------------------ | | **Reports** | List your reports, read a report's chapters, sections, fields and observations | | **Chapters & sections** | Add, rename, remove and restore chapters; clone and reorder sections | | **Fields** | Read and update data fields and their options | | **Observations** | Write, rewrite, retag and move observations between sections | | **Library** | Create, edit, merge, duplicate and reorganize library categories and entries | | **Tags & Smart Tags** | Create and edit tag groups, fix up tags across a report after a merge | | **Boilerplate** | Read and edit your boilerplate language | | **Summary tables** | Edit the tables that appear in the report overview | | **Photos** | List the gallery, look at a photo, and edit captions and tags | | **PDF forms** | Read and edit your PDF form field mappings | | **Report settings** | Change template and report-level options | It cannot publish reports, take payments, touch your schedule, or see anything belonging to another company. ## It Knows What You're Looking At This is the part that makes it feel less like an API and more like an assistant. ScribeWare tells the AI which company you're logged into, which report is open, and which chapter you're on. So you never have to give it an ID, a report number, or a file. *"Fix the typos in this chapter"* is a complete instruction — it already knows which chapter you mean. ## What You Need * **The ScribeWare desktop app, running and logged in.** The AI talks to ScribeWare through your own computer. Close the app and the connection goes away; log out and it stops. * **An AI tool that supports MCP** — Claude Desktop, Claude Code, Gemini CLI and most others do. See [Connecting Your AI Tool](/into-the-weeds/ai-assistant-mcp/connecting-your-ai-tool.md). * Nothing on the mobile app. This is desktop only. ## Is My Data Going Anywhere? The connection itself is entirely local — ScribeWare listens on your own machine, at an address only your machine can reach. Nothing is opened up to the internet, and no ScribeWare data is sent anywhere new by turning this on. What *does* leave your computer is whatever your AI tool sends to its own provider, which is the same arrangement as any other use of that tool. If you ask Claude to rewrite an observation, that observation text goes to Anthropic, exactly as it would if you'd pasted it into the chat window yourself. Your choice of AI tool decides where your report content goes, so pick one whose terms you're happy with. ## Where to Next {% content-ref url="/pages/Fr0iDLAX4Or9CYei9Wy9" %} [Connecting Your AI Tool](/into-the-weeds/ai-assistant-mcp/connecting-your-ai-tool.md) {% endcontent-ref %} {% content-ref url="/pages/Fo3BGc18QKtsRwEcelGY" %} [Teaching It Your Style](/into-the-weeds/ai-assistant-mcp/teaching-it-your-style.md) {% endcontent-ref %} {% content-ref url="/pages/j5Aa3CzXQaMwFdpyG4wz" %} [Reviewing AI Changes](/into-the-weeds/ai-assistant-mcp/reviewing-ai-changes.md) {% endcontent-ref %} {% content-ref url="/pages/CvOATTLJpwPEEXqsIc42" %} [What You Can Ask For](/into-the-weeds/ai-assistant-mcp/what-you-can-ask-for.md) {% endcontent-ref %} {% content-ref url="/pages/BCnJsrWSTeNbyhp9SinY" %} [Troubleshooting the AI Assistant](/into-the-weeds/ai-assistant-mcp/troubleshooting-the-ai-assistant.md) {% endcontent-ref %} --- # 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 dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.scribeware.com/into-the-weeds/ai-assistant-mcp.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. 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.