Model Context Protocol (MCP) was all the rage in the tech world in 2025 and will likely stay that way throughout 2026. MCP is changing how developers bridge the gap between AI applications and local or remote data. One of these bridges is the MCP server, which exposes capabilities to AI applications through tools, resources, and prompts.

As developers are the primary users of MCP servers, companies need to think carefully about developer experience (DX) when releasing one. For most developers, that experience starts with the documentation.

So today, we are outlining what makes for great MCP server documentation. We have highlighted five real-world examples of excellent MCP server documentation from the following companies: Figma, GitLab, PandaDoc, SmartBear, and Webflow.

Components of Good MCP Server Documentation

As more companies build and release MCP servers, patterns have emerged that are indicative of quality MCP server documentation:

• Main menu that is well-organized and easy to navigate.
• Prominent links to relevant pages and information.
• Clear instructions for MCP server installation and configuration.
• Tools organized into use-case-driven sections.
• Explanations of what the tools are and what they accomplish.
• Examples of good natural language prompts for guiding the AI.
• Troubleshooting or FAQ section covering common issues and fixes.

It is critical for companies that offer an MCP server to provide documentation as outlined because it:

• Helps developers and AI applications better understand what the server does and the tools it offers.
• Makes it easier for developers to configure the server and connect to the AI, preventing frustration that could lead them to look for well-documented alternatives from competitors.
• Shows the possibilities the MCP server offers through use cases and capabilities, encouraging developers to innovate when building AI applications.
• Improves the developer experience, which in turn can lead to increased adoption of the server and a growing community around it and the platform.

Now, let’s look at some examples of excellent MCP server documentation. We chose these because they go well beyond a GitHub README, and most of them meet all the criteria outlined above.

Figma MCP Server Documentation

Website | MCP documentation

Figma is a collaborative design platform that enables teams to work together on brainstorming, planning, designing, and developing various digital projects.

The company announced the beta release of its official local MCP server in June 2025. The server allows developers to access context from Figma and bring it into AI tools such as Cursor, Windsurf, and Claude Code. For example, an AI agent could generate code from a selected Figma frame, or an IDE could pull context from designs in Figma.

What We Like About the Documentation

Figma has created documentation with a simple design that is easy to navigate. Everything you need is listed on the main navigation menu on the left. One of the best parts of the documentation is the “tools and prompts” section. It clearly explains the different tools the Figma MCP server provides, along with some examples of suggested prompts.

The Q&A on the left menu is quite helpful for developers who have run into issues when using the server. Another great feature is the “how to get the best output” section in the navigation menu, which gives developers a better chance of success.

Conclusion: Figma’s documentation gives developers a big head start when working with the MCP server.

GitLab MCP Server Documentation

Website | MCP documentation

GitLab is a comprehensive DevSecOps platform that allows teams to complete numerous tasks in various categories. These categories include source code management, continuous integration, security, and continuous delivery.

The company announced support for Model Context Protocol in September 2025, which includes an MCP server that enables AI tools to securely connect to a GitLab instance. The server is available through the GitLab Duo add-on (Core, Pro, or Enterprise), and it currently supports these use cases: issues, merge requests, pipelines, and server info.

What We Like About the Documentation

GitLab’s MCP server documentation is easy to navigate and provides clear and detailed installation instructions for a number of AI tools, including Cursor, Claude Code, and Gemini Code Assist. There is a troubleshooting section that explains several issues that could occur while working with the server. It would be helpful to have more common (and not-so-common) issues covered in this section.

The best part of GitLab’s MCP server documentation is the “Server Tools” section, which outlines all the tools available along with natural language examples. The tools are not organized into use-case-driven sections, though, which could be a nice structural change to improve navigation.

Conclusion: GitLab’s documentation gets the job done.

PandaDoc MCP Server Documentation

Website | MCP documentation

PandaDoc is a document automation software platform that you can use for a variety of paperless business use cases, such as obtaining eSignatures and managing contracts.

The company announced the release of its new MCP server in September 2025. It allows developers to build AI agents that drive complete agreement flows. For example, an AI agent could leverage PandaDoc through the MCP server to instantly assemble contracts or handle dynamic processes like compliance checks.

What We Like About the Documentation

The PandaDoc documentation is designed simply, with straightforward navigation that makes it easy to find the information you need to get the server up and running. One standout feature of the documentation is its prompt library. It’s chock full of prompt examples for many use cases — create documents, track status, search and filter, and the list goes on.

There is a section covering authentication and permissions, which is helpful. However, it would be great to see more thorough guidance on authentication and additional security best practices. There is also a “Troubleshooting” section, although it seems light on content. If developers have encountered other issues with the server, it would be helpful to cover them in this section.

Conclusion: PandaDoc is an excellent example of effective MCP server documentation. It checks all the boxes.

SmartBear MCP Server Documentation

Website | MCP documentation

SmartBear provides a suite of testing and observation products that help developers ensure quality throughout the entire API and software lifecycle.

Released in August 2025, SmartBear’s official MCP server allows developers to access the context and capabilities across the company’s products using natural language prompts initiated by AI assistants and IDEs. For example, a developer could build an AI agent that answers questions about BugSnag errors or active QMetry tests.

What We Like About the Documentation

SmartBear’s documentation is very easy to navigate (there are only four sections). The “Getting Started” section includes clear installation and setup instructions, MCP host configuration examples, and usage examples. The “MCP Server Capabilities” section provides a list of available tools for each SmartBear product.

The comprehensiveness of the tool lists is great. Each list provides a description of each tool that includes its purpose, parameters (when applicable), what it returns, and its use case. There are also examples of natural language queries for each tool. These tool lists help developers and the AI understand what tools are available for each SmartBear product and when to use them.

The documentation includes a “Best Practices” section, although more guidance on security considerations would be nice. Bonus points for the nice illustration on the main page. We looked at dozens of MCP server documentation sites, and most are solid text with no graphics.

Conclusion: SmartBear’s documentation gives you a solid foundation for working with the MCP server.

Webflow MCP Server Documentation

Website | MCP documentation

Webflow is a website experience platform that provides tools for building, managing, and optimizing custom responsive websites.

Released in April 2025, the official Webflow MCP server exposes Webflow’s APIs as tools that AI agents can use. Available tools fall into two categories: Data API tools and Designer API tools. The server lets you prompt an AI agent to perform tasks in Webflow, such as updating a design, managing site data, or creating elements.

What We Like About the Documentation

Like the others on this list, Webflow’s MCP server documentation is well-designed and easy to navigate. The graphic on the main page is a nice touch.

Webflow’s documentation has excellent attention to detail regarding the sections on available tools and the prompt library. Each tool page explains in detail what the tool is, what it accomplishes, and includes an example of a returned response.

The prompt library is phenomenal — every prompt is “ready-to-use” and has a step-by-step breakdown of how it works. Developers can use these as is or as a reference for creating their own. Currently, the library only has prompts for Designer API tools. Hopefully, prompts for Data API tools will be added soon.

There is a solid FAQ section that includes questions and answers about troubleshooting issues. This section can help developers solve common issues when implementing the server.

Conclusion: Webflow’s MCP server documentation does not just check the boxes, it smashes them.

MCP Server Documentation: Room For Improvement

The five examples we have highlighted meet or exceed the current criteria for excellent MCP server documentation. However, other attributes could improve them even more:

• Detailed guides: Developers would have an even greater chance of success if MCP server documentation also included guides that explain step-by-step common operations and use cases for the MCP server.
• Guidance on authentication: Some of the documentation sites we reviewed cover authentication, but not extensively. It would help developers immensely if documentation included thorough guidance on authentication. Authentication is tricky to implement, especially for AI agents.
• Best practices: Some of the documentation we have seen has a best practices section, but it is light on content. It would be great for documentation to have a comprehensive list of best practices, especially for security.

It will be interesting to see how the industry shapes best practices for creating excellent MCP server documentation over time. In the meantime, the current criteria outlined above is a good starting point for companies offering an MCP server for external developers.