Document360

Document360: Features And Certifications

Document360: Knowledge Bases With Eddy AI

• AI search that cites every source, so an answer can be checked against the article behind it.
• An MCP server exposing your knowledge base to ChatGPT, Claude and Copilot rather than trapping it in a portal.
• Migration paths from Confluence, Zendesk, SharePoint and Google Docs, which is most of the market's incumbent.
• Enterprise controls throughout — SSO, SCIM, audit logs, IP restriction, role-based access, workspaces.
• SOC 2 and ISO 27001:2022, with 60-plus out-of-the-box connectors.

What Does An MCP Server Add To Documentation?

Document360 lists an MCP server among its core AI features, described as connecting your knowledge base to ChatGPT, Claude and Copilot.

That is a more consequential decision than it appears, and it turns on where people now go for answers.

A knowledge base has always assumed the reader comes to it. They open the help site, search, and read an article. That assumption held for twenty years and is quietly breaking, because people increasingly ask an assistant instead — and the assistant answers from whatever it can reach.

If your documentation is not reachable, the assistant answers anyway, from training data that may be years old or from a competitor's page. You do not get a blank; you get a confidently wrong answer with your product's name on it.

An MCP server closes that gap by making the knowledge base a source an agent can query directly. It is the same structural move that makes personal document assistants like Khoj useful, applied to a company's public documentation rather than an individual's files.

A knowledge base assumes the reader comes to it. People now ask an assistant instead, and the assistant answers whether or not it can reach you.

Search That Cites Its Own Articles

Document360 describes its AI search as conversational search grounded in your articles that cites every source.

The citation is the whole feature, and skipping it is what makes most documentation chatbots unusable in practice.

An uncited answer in a support context creates a specific problem: the reader cannot tell whether it came from current documentation, an outdated article nobody archived, or the model filling a gap. All three read identically, and only one is safe to act on.

That matters more here than in general search because documentation answers get acted on. Someone follows the steps, changes a setting, and finds out later whether the answer was real.

Citations also change the maintenance loop in a useful direction. When an answer is wrong, the citation tells you which article to fix — so the AI layer becomes a way of finding stale documentation rather than a way of hiding it.

Migration Away From Confluence And Zendesk

The migration pitch is direct: most documentation tools were built for a world before AI agents, MCP and connected workflows, and named import paths exist from Confluence, Zendesk, Microsoft Word, SharePoint and Google Docs.

Listing those five is a precise piece of positioning, because between them they hold most of the world's existing internal documentation.

The argument is honest about the real obstacle, too. Nobody stays on a documentation platform because they love it; they stay because moving several thousand articles is a project nobody wants to sponsor.

Naming the import paths reduces that objection from a project to a task, which is the only way this category's incumbents ever get displaced.

The caution is the one every migration carries: articles move, but structure, redirects, permissions and internal links often do not. A knowledge base's URLs are load-bearing — they sit in support macros, in product tooltips and in search results — and the question to ask a vendor is what happens to every existing link on the day you switch.

• Ask what happens to existing URLs and redirects during migration.
• Test AI search citations against articles you know are outdated.
• Check the MCP server against the assistants your customers actually use.
• Map role-based access before importing anything internal.
• Use the analytics to find the articles support keeps linking manually.

The Support Ticket Deflector And What It Measures

Among the 100-plus features sits a support ticket deflector, alongside pro analytics, duplicate content detection and a feedback manager.

Deflection is the metric this entire category is bought on, and it deserves a careful reading.

The theory is clean: a customer who finds the answer in documentation does not open a ticket, so tickets avoided is documentation value delivered. One of the platform's published case studies quotes a 15 percent reduction in support tickets after localising a knowledge base.

The complication is that not opening a ticket is not the same as being helped. A customer who searches, fails, and gives up also does not open a ticket, and looks identical in the deflection number.

Which is why the feedback manager and duplicate content detection matter more than they sound. Deflection tells you tickets went down; article-level feedback and search-with-no-result data tell you whether that happened because people were helped or because they left. Reading what customers actually said across every channel is a further step again, and where Thematic operates.

Reading The Ratings And The Case Studies

Document360 displays 4.7 out of 5 on G2 and describes itself as a G2 leader in knowledge base software for three consecutive years. The review count is given as 1,400-plus in one place and 3,000-plus in the footer.

Two different counts on one page usually means two different scopes — one platform versus several, or verified versus total — and neither is labelled, so read the score rather than the volume.

The case studies are more useful because each names a company, a person and a specific outcome: 10,000-plus page views on individual articles, a 30 percent traffic increase after analytics-led content changes, a 15 percent ticket reduction after localisation.

Attributed outcomes with named people are checkable in a way an aggregate rating is not, and the localisation result in particular describes a mechanism you can reason about.

Three years of category leadership is the more durable signal. Any product can have a strong quarter; sustaining a position across a rolling review base is harder to arrange. For knowledge that stays internal rather than customer-facing, TextCortex is closer to the right shape.

Launching A Knowledge Base People Read

Documentation projects fail in the same way twice: everything is migrated and nothing is maintained, or everything is rewritten and nothing ships.

This sequence puts the decisions that are expensive to reverse first.

1. Settle URL structure and redirects before importing a single article.
2. Import one category rather than the whole estate, and check what breaks.
3. Turn on AI search and test it against articles you know are stale.
4. Configure role-based access before any internal content lands.
5. Connect the MCP server and ask an assistant a real customer question.
6. Review search-with-no-result data monthly and write against it.