Hello, this is 3Rabbitz.
Writing with AI is already familiar. But managing a product manual of hundreds of pages together with AI is still hard. You can't copy and paste a whole document, and moving AI's edits back into the original document costs time and puts quality at risk. Above all, an AI that doesn't know the document structure can break formatting or edit the wrong place.
ManualWorks solves this problem with MCP (Model Context Protocol). Claude Code, Codex, and Gemini CLI read the ManualWorks document structure directly, and search, proofread, and add content chapter by chapter and element by element. Because AI works right inside the document without copying and pasting, product and technical writing teams can keep their documents accurate, faster.
Here are some examples of what changes in real workflows.
Exploring documents: find what you need right away
Case 1. Answer customer questions right away
A customer writes, "I get an authentication error when calling the API." You can't remember offhand which chapter of which document covers it.
Conversation with AI
User: Search the documents for "API authentication error."
AI: I found related content in 2 documents.
1. [API Reference] Chapter 4 "Authentication" — How to get an API key and set the authentication header
2. [Troubleshooting Guide] Chapter 2 "Common Errors" — Causes and fixes for 401/403 errors
User: Read chapter 2.
AI: Here is chapter 2, "Common Errors."
- 401 Unauthorized: The API key has expired or is invalid...
- 403 Forbidden: You don't have permission for the resource...
With a single keyword, you find where the related content is, without opening several documents yourself. You spend less time on customer inquiries, and your answers are better grounded.
Case 2. Find things without knowing the exact term
A new team member asks, "Is there a way to limit which pages each user can see?" They don't know whether the feature is called "access control," "permission settings," or "role management."
Conversation with AI
User: Do a semantic search for "how to set different pages each user can access."
AI: I found related content in 3 documents.
1. [Administrator Guide] Chapter 3 "Role-Based Access Control" — Setting menu access permissions by role
2. [Administrator Guide] Chapter 5 "Managing User Groups" — Restricting page access by group
3. [Operations Guide] Chapter 2 "Initial Setup" — Default roles and permissions
User: Read chapter 1.
AI: Here is chapter 3, "Role-Based Access Control"...
Semantic search finds content by meaning, not exact terms. Even without knowing the words "access control," you can find related documents from a description alone, so new members can catch up on product knowledge faster.
Semantic search works only when the chatbot is set up.
Case 3. Check specs in the documents while writing code
You need the list of error codes during development. You know it's in the documents, but opening a browser to look for it is a hassle.
Conversation with AI
User: Show me my documents.
AI: Here are the 5 documents you can access.
1. API Reference [a1b2c3...]
2. User Guide [d4e5f6...]
3. Operations Guide [g7h8i9...]
...
User: Search the API Reference for the keyword "error code."
AI: I found 2 results in chapter 5, "Error Handling."
- Element 1: Error code system (explanation of the 1xx to 5xx ranges)
- Element 2: Detailed error code table (code, message, action)
User: Show me element 2.
AI: Here is the detailed error code table.
| Code | Message | Action |
| 1001 | Invalid parameter | Check the request parameter format |
| 1002 | Missing required field | Check for missing required fields |
...
You look up documents right inside your development tools. You don't lose context switching between a browser and an editor, so checking specs and referring to documents during implementation gets faster.
Case 4. Find the impact before changing a feature
You're about to change the response format of an existing API. You need to know in advance which documents mention this API.
Conversation with AI
User: Search all documents for "getUserList."
AI: I found related content in 3 documents.
1. [API Reference] Chapter 3 "User API" — Description of the getUserList endpoint (2 results)
2. [Developer Guide] Chapter 7 "User Integration" — getUserList call example (1 result)
3. [Release Notes] v2.3 — email field added to the getUserList response (1 result)
User: Read chapter 1.
AI: Here is chapter 3, "User API." getUserList is described in the following elements...
With a single API name, you find every related document at once. By knowing in advance which documents need updating after the change, you can avoid missing documents right before a release.
Writing documents: from proofreading to adding content
Case 5. Proofread a 200-page manual
Before a product launch, you need to review the entire manual. Reading more than 200 pages line by line to find typos isn't realistic.
Conversation with AI
User: Read the following document. [document ID]
AI: The document has 15 chapters.
User: Read and proofread all chapters in order. Ask me to confirm before applying changes.
AI: I've checked chapter 1, "Getting Started." There are 3 things to fix.
1. "After complete the installation" → "After completing the installation"
2. "e-mail" → "email" (consistent terminology)
3. "Check it.." → "Check it." (duplicate period)
User: Fix them.
AI: I've finished fixing chapter 1. I'll check chapter 2...
Proofreading that would take a person days is processed chapter by chapter. Edited elements show up in the editor in real time, and the @mcp tag sets AI edits apart in the history, so reviewing and approving is clear, too.
Case 6. Write documentation for a new feature through conversation
A new feature has been added. You need to put its description in the right place in the existing manual.
Conversation with AI
User: Read the following chapter. [chapter ID]
AI: This is the "Exporting Data" chapter. It currently describes CSV and JSON export.
User: Add an "Excel Export" section after JSON export.
- Supports exporting to xlsx
- The sheet name follows the document title
- Only table elements can be exported
AI: I'll add the following.
## Excel Export
Exports data in xlsx format...
Shall I add it?
User: Add it.
AI: Done.
User: Change "You can export" to "Export."
AI: Changed.
You shape the content through conversation and apply it right where you want in the existing manual. Drafting, placing, and repeated revisions happen in one flow, so documenting new features takes less time.
Case 7. Unify terms across documents at once
Because different people wrote them, the same concept appears as "delete," "remove," and "erase."
Conversation with AI
User: Read the following document. [document ID]
AI: The document has 12 chapters.
User: Change every "remove" and "erase" to "delete" in all chapters.
AI: I've found the elements that need changes.
- Chapter 3: "Remove the item" → "Delete the item" (2 cases)
- Chapter 5: "Before erasing the data" → "Before deleting the data" (1 case)
- Chapter 8: "Remove the existing settings and" → "Delete the existing settings and" (3 cases)
6 cases in total. Shall I change them?
User: Change them.
AI: I've changed all 6 cases.
AI considers the context and changes words only where it makes sense. It cleans up terminology more safely than a simple find and replace, improving the quality of wording across your documents and the consistency of your product message.
Keeping quality consistent: style guides
If AI writes in a different tone every time, the consistency of your brand and product descriptions suffers. To prevent this, ManualWorks supports two style guides.
ManualWorks document format — Rules for handling document structure and elements correctly
Writing style guide — Rules for keeping the tone of content consistent, such as voice, terminology, and sentence length
Save the style guides in your AI's memory, and it works by the same standards even in new sessions. Voice, terminology, and structure stay consistent even when the people in charge change.
Getting started
You can start with MCP by creating an API key and doing some simple setup. Claude Code, Codex, and Gemini CLI are all supported, so you can work with ManualWorks documents right in the AI tool you already use. For detailed setup, see the Using ManualWorks in Claude Code/Codex/Gemini CLI guide.
If you already use ManualWorks, speed up searching, proofreading, and updating your documents with MCP. If you don't use ManualWorks yet, see for yourself how product documentation changes in the age of AI.