As of October 29, 2026, at least ten product adoption tools run an MCP (Model Context Protocol) server that lets Claude, ChatGPT, Cursor and other AI assistants work with your onboarding content. They differ in one way that matters: some let the assistant build tours and other in-app guidance as drafts, while others only let it read analytics and manage segments.
Cueston makes one of these tools, so read this with that in mind. Every claim about another product below comes from that vendor's own documentation, linked in each section. If a vendor's page doesn't say something, we don't claim it. These servers change fast: Appcues and Userflow updated their MCP docs on October 26 and 27, 2026.
The short answer
- The assistant can build in-app guidance as drafts: Appcues (beta), Userflow, Usertour (beta), Chameleon, Clew and Cueston.
- Mostly read and analyze, with some writes: Pendo (segments, guide state and audience), Userpilot (segments, reports, survey drafts) and UserGuiding (segments, users, help articles and product updates).
- Read-only: Product Fruits.
What to check before you connect one
- Can it create the content types you use (tours, tooltips, banners, checklists, surveys), or only some of them?
- Does what it builds stay a draft until a person publishes it? Can you require review?
- Can it check its work against your real pages, or does it guess CSS selectors?
- Can you choose its access level (read only, drafts, publish) when you connect?
- Does every action show up in an audit log under the AI app's name?
Appcues
Appcues's hosted MCP server can build experiences in beta: "Your AI client can create and edit Flows 2.0, Embeds, Banners, and Pins as drafts," and "nothing goes live until you publish, or ask your AI client to publish." An admin turns it on under Settings → Appcues AI, and each connection has the permissions of the person's role. Source: Appcues MCP Server, updated October 27, 2026.
Userflow
Userflow's MCP reads flows, users, segments and analytics, and "can also build and change things in your Userflow account." Flows are always saved as drafts, and publishing "is a separate step that asks you to confirm every time." Segments, charts and dashboards are created immediately. Checklist, launcher and resource center content is still added in the app. Source: Userflow MCP, updated October 26, 2026.
Chameleon
Chameleon connects MCP clients to Chameleon Copilot, which lets you "use natural language to build, edit, and analyze your Chameleon Experiences." It is available on all plans. The guide doesn't list the individual actions. Source: Chameleon MCP user guide.
Usertour
Usertour, an open-source tool, labels its MCP endpoint beta. With write scopes it can create, update, duplicate and delete content, manage versions and translations, validate a version and publish or unpublish it. Source: Usertour MCP.
Clew
Clew lists an "MCP server: manage tours from Claude, Cursor and ChatGPT" on its Pro plan ($14.99 a month) and above; the free plan doesn't include it. Its pricing page doesn't detail the tools. Source: tryclew.io.
Pendo
Pendo's MCP is mostly read-only. Optional write tools, which an admin enables, create and update segments, change a guide's state (draft, staged, public or inactive) and its audience, manage feedback and ideas, and work with Orchestrate journeys. The tool list has no tool for writing a guide's steps or content. Source: Pendo MCP tools for AI agents, updated October 26, 2026.
Userpilot
Userpilot's MCP server creates and edits segments, reports and dashboards, and creates survey drafts. Its docs say it can't "build or publish in-app content such as flows and banners" and can't delete anything. It is available on all plans. Source: Userpilot MCP server.
UserGuiding
UserGuiding's MCP server has 62 tools on all plans. It writes users, segments, knowledge base articles and product update posts (saved as drafts), and approves roadmap requests. For guides, checklists and hotspots the documented tools are analytics, such as step funnels and checklist stats. Source: UserGuiding MCP Server.
Product Fruits
"Product Fruits MCP is currently read-only." The assistant "cannot create, edit, publish, unpublish, or delete Product Fruits content." Source: Product Fruits MCP Connection.
Cueston
Cueston's MCP server has 47 tools and works on every plan, at one URL for Claude, ChatGPT, Cursor, VS Code and other MCP apps. The assistant can build tours, tooltips, banners, checklists, surveys and announcements as drafts. Before it suggests a selector, it checks the element on pages Cueston has seen on your site. It then validates the draft and sends you a preview link that opens on your own product.
- When you connect, you pick one access level: read only, create and edit drafts, or edit and publish.
- If your workspace requires approval, the assistant requests review instead of publishing.
- It can roll a guide back to an earlier version, translate content, and read analytics, survey results and broken targets.
- Every call is in the audit log under the name of the app that made it.
Setup takes a minute: connect Cueston to Claude, ChatGPT or Cursor. The full tool list is in the MCP documentation.
Frequently asked questions
What is an MCP server in a product adoption tool?▾
MCP (Model Context Protocol) is an open standard that lets AI assistants such as Claude, ChatGPT and Cursor call a product's tools. A product adoption tool's MCP server lets the assistant read your onboarding analytics and, depending on the vendor, build or change tours, tooltips and other in-app guidance.
Which onboarding tools let an AI assistant build product tours?▾
As of October 29, 2026, the vendors' docs say Appcues (beta), Userflow, Usertour (beta), Chameleon, Clew and Cueston let an assistant build or edit in-app guidance through MCP. Pendo, Userpilot and UserGuiding document no tool for writing guide content, and Product Fruits is read-only.
Can the assistant publish without a person checking?▾
It depends on the tool and your settings. Appcues, Userflow and Cueston save what the assistant builds as drafts. Userflow asks for confirmation before every publish. In Cueston you choose whether the assistant may publish at all, and workspaces with approval rules send its work to review.