Feature Request

Support for documentation-wide exact full-text search, batch article retrieval, and machine-readable export in the Document360 MCP app. This will enable efficient searches across large knowledge bases without having to retrieve articles individually. Problem statements Documentation-wide exact full-text search — Problem: the search tool is semantic-only, so a lookup for a known literal token (his placeholder tag, a code string, a product term) silently fails across the KB — the exact experiment that broke on the call. Solution: add an exact/literal (and regex) match mode to search that scans full article text across every category and version and returns all occurrences Return a scoped match list with URLs — Problem: his QA workflows start from a list, scoped to a whole category or the entire KB — not one article — and today no tool produces one. Solution: have search (or a new list-articles) return a structured list of matches — ID, title, URL, category/version metadata — with category/KB scope filters and a result-count limit, exactly the shape confirmed on the call. Currently Mintlify and Gitbook MCP server tools have solution to above problem statements. Please add this requirement to your MCP server tools backlog and prioritise. References https://www.mintlify.com/docs/ai/model-context-protocol#mcp-tools https://gitbook.com/docs/developers/gitbook-api/api-reference/spaces/space-content/full-text-search-across-all-pages-in-a-space

When I fetch an article through the Document360 MCP get-article tool, snippets and variables come back as raw placeholders like {{ snippet.NAME }} and {{ variable.NAME }} instead of their actual values. I'd like the MCP to return the full article content with snippet and variable values included, the same way the editor preview and published pages render them.

Add support for the isForDisplay parameter (or an equivalent option) to the MCP get-article so it can return fully rendered article content with snippets, variables, and glossary terms resolved. Currently, MCP returns raw content with unresolved placeholders such as {{ snippet.NAME }} and {{ variable.NAME }}, requiring a separate REST API call to retrieve the rendered version. Supporting this in MCP would simplify integrations and provide output consistent with the article preview and published content. Thank you. Thank you.

The current MCP Server Analytics dashboard provides a solid baseline, but one of our customers feels that there are several gaps that limit its usefulness for admins managing real-world deployments. They would like to request the following improvements: Break down MCP usage by user role (Owner, Admin, Editor, etc.) and distinguish Reader vs Team Member activity separately, so admins can understand who is driving usage and plan enablement accordingly. Add a geographic and timezone distribution view to help global teams correlate failure spikes or usage patterns with specific regions. Surface which articles and categories are being accessed through MCP tools — not just which tools are called — to identify content gaps and validate whether AI assistants are reaching the right information. For multi-project accounts, introduce a workspace-level filter or cross-project summary so admins can compare MCP activity without switching between projects. Extend export beyond PNG chart images to full CSV downloads across call logs, tool usage, and user activity, enabling use in external reporting and BI tools. Add a "Compare to previous period" toggle so adoption trends can be tracked week-over-week or month-over-month directly within the dashboard. Surface token consumption per user and per client alongside call volume so admins can proactively manage plan limits before failures occur.

I can think of multiple scenarios where AI-based cleanup tools could do the heavy lifting for labor-intensive maintenance tasks: 1) Drive: scan files and compare them against KB structure; identify files that are obsolete, that appear to be stored in the wrong folder, that have meaningless file names etc; offer to make changes directly in agent mode (including creating categories and moving/deleting files). 2) Content Reuse: Because MCP cannot access content reuse items, an AI-based cleanup tool could scan content reuse items like snippets and look for redundancies, duplication, gaps etc.; to make it even more powerful, instead of just scanning and comparing the items themselves, it could also compare these items against the KB. Especially for snippets, this tool could also help with the basic task of searching snippet texts, which would be a major improvement to searching only snippet names.

The customer is building an AI-powered chat interface using the Document360 MCP server as the knowledge backend. The interface is intended to serve multiple end users who are not Document360 project members. The current OAuth flow requires each user to have a Document360 account and active project membership, which blocks this use case entirely. Customer's ask is the support for secret key / API token based authentication on the MCP server, allowing their backend service to connect to MCP without per-user login.

I’m using the Document360 MCP to retrieve and update Knowledge Base articles, and I’m running into an issue where long article bodies are not always returned in complete, usable form. I’m not sure whether the content is being truncated by the MCP server itself or collapsed somewhere between the MCP response and the client, but the practical effect is the same: I request an article, and the body I receive may be incomplete. This creates a problem for write workflows because article updates appear to require sending back the complete article body, rather than applying a small targeted patch. A typical flow looks like this: I call get_article for an article. The response includes metadata and body content, but a long middle section of the body may be omitted or collapsed. I need to make a targeted edit to one section. To save the update, I need to call update_article with the revised article body. Because I don’t have the complete original body, I either have to reconstruct the missing section or risk unintentionally changing/removing content outside the intended edit. After saving, I call get_article again to verify, but if the readback is also incomplete, I still can’t fully confirm that the article body was preserved. So the core issue is not just that long articles are harder to review. It’s that incomplete readback makes full-body replacement updates risky. If update_article requires replacing the full body, then get_article needs to provide the full body reliably. A few improvements would help a lot: Return the complete article body from get_article . If full bodies are too large, provide paginated or chunked body retrieval. Include an explicit is_truncated flag when content has been shortened. Support targeted patch-style updates for smaller edits. Preserve and expose the article’s content/editor type consistently during read and update operations. The MCP is very useful for documentation maintenance, but safe editing depends on being able to retrieve, modify, save, and verify the complete article content.

Connect Feedback Manager details to MCP so I can use other tools to run analytics and understand the feedback + create a prioritization plan for necessary updates

When Document360 is connected to Microsoft Teams via MCP, the connector currently returns text-based content from articles. We would like the MCP connector to also return images embedded within articles as part of its responses so that users in Teams receive the full context of the article content, not just the text. This is particularly valuable for visual documentation such as step-by-step guides, diagrams, and annotated screenshots.

It would be great if we could analyze comments with MCP. We use comments for proofreading. With MCP, we could analyze proofreading comments for patterns and issues that occur frequently, and then derive styleguide skills from it.