The DYNO Mapper MCP Server Is Now Hosted: 108 Tools, No Install

What changed

Three things are different from the original launch.

It runs on our servers, not yours. The original server was an npm package, @dynomapper/mcp-server, that ran on your own machine. The hosted server lives at mcp.dynomapper.com, so any AI client that supports remote MCP servers can reach it.

You sign in with your DYNO Mapper account. The hosted server uses OAuth. When you add it to your client, a DYNO Mapper sign-in window opens, you approve access, and you’re done. There’s no Personal Access Token sitting in a config file.

Agents can now change things, not just read them. The first release was mostly read-only: agents pulled audits, scores, rankings, and inventories. Of the 108 tools today, 42 create, change, or delete data, from sitemap restructures to comments to keyword tracking.

The npm package is still available if you’d rather run the server locally. The MCP server documentation covers both options.

What it means for DYNO Mapper users

Setup takes about a minute

The old setup assumed you were comfortable with a terminal and a JSON config file. That’s fine for developers, but it kept a lot of SEO managers, content strategists, and accessibility consultants from trying the MCP at all. With the hosted server, adding DYNO Mapper to Claude looks like adding any other connector: paste a URL and sign in.

It works in more places

A local server needs a machine that can run Node.js. A hosted server doesn’t, so DYNO Mapper now works in browser-based clients such as Claude on the web, as well as in Claude Desktop, Claude Code, Cursor, VS Code, and other clients that support remote MCP servers with OAuth.

You always have the latest tools

When we add or improve a tool, the hosted server picks it up right away. There’s no package to update and no version drift between teammates.

Easier rollout across a team

Each person signs in with their own DYNO Mapper account, so the agent sees what that person can see. There are no shared tokens to hand around and no config files to keep in sync.

The 33 new tools

Sitemap editing (15 tools)

This is the biggest addition. Agents can now build and restructure sitemaps the same way you would in the sitemap editor: add pages and containers, rename and retag them, move and sort them, duplicate sections, and read or replace a page’s content blocks. Bulk tools handle up to 500 items in one call.

Every edit goes into the project’s undo history, which is shared with the DYNO Mapper app. If an agent makes a change you don’t like, you can undo it in the editor, or ask the agent to undo it.

“In our redesign project, create a Resources section, move every guide and whitepaper under it, tag them ‘resources’, and mark the outdated ones for consolidation.”

Tools: create_project_sitemap_item, update_project_sitemap_item, delete_project_sitemap_item, duplicate_project_sitemap_item, move_project_sitemap_item, sort_project_sitemap_items, bulk_update_project_sitemap_items, bulk_move_project_sitemap_items, bulk_delete_project_sitemap_items, get_project_sitemap_item_content, replace_project_sitemap_item_content, undo_project_sitemap_edit, redo_project_sitemap_edit, get_project_sitemap_undo, and get_project_sitemap_redo.

SharePoint crawling and sign-in prompts (5 tools)

Agents can now crawl a private SharePoint team site with create_project_from_sharepoint. If the account uses MFA, the crawl pauses and waits for the code; the agent finds the waiting prompt with get_project_login_inputs and sends your code with submit_project_login_input. Accessibility runs on sites behind a login work the same way, with get_accessibility_login_inputs and submit_accessibility_login_input. Prompts expire about an hour after they appear.

“Crawl our intranet at mycompany.sharepoint.com, then run a WCAG 2.2 AA check on it and tell me the three guidelines with the most problems.”

DYNO Mapper never stores the SharePoint username and password. Because they’re passed to the tool, though, they will appear in your AI conversation, so use a dedicated crawl account.

D.A.C.Q. scores (1 tool)

Private SharePoint sites have no search visibility to measure, so they’re scored with D.A.C.Q. (Discoverability, Accessibility, and Content Quality) instead of V.A.C.Q. The new get_project_dacq_score tool reads it. Like V.A.C.Q., the overall score appears once every pillar has been measured, which means running an accessibility check first.

Categories and workflow statuses (8 tools)

Agents can create, rename, recolour, reorder, and remove the categories and workflow statuses you use to label sitemap pages. Combined with the bulk sitemap tools, that means an agent can set up a content audit’s labels and apply them across hundreds of pages in one conversation.

Tools: create_project_category, update_project_category, delete_project_category, move_project_category, create_project_workflow_status, update_project_workflow_status, delete_project_workflow_status, and move_project_workflow_status.

Comments (4 tools)

Agents can now post, edit, and delete comments on sitemap items, and read every comment in a project at once with get_all_project_comments. That makes it easy to leave review notes on the pages an audit flagged, or to summarize open feedback before a client meeting.

“Find every page in project 412 with a missing meta description and leave a comment on each one asking the content owner to add it.”

Tools: create_item_comment, update_item_comment, delete_item_comment, and get_all_project_comments.

Good to know

• Keep approvals on for write tools. Most AI clients can ask before a tool runs. Since agents can now change your sitemaps and comments, it’s worth leaving that on for tools that change data.
• Some deletes are permanent. Sitemap edits can be undone, but deleted comments and deleted accessibility runs can’t be restored through the MCP server.
• SEO audits still start in the app. The audit tools read the results of your latest audit; they don’t start a new one.
• Scores need their inputs. A V.A.C.Q. score needs a completed accessibility run and tracked keywords. A project that has only been crawled will show empty pillars.

Get connected

Claude (web and desktop): open Settings → Connectors, choose Add custom connector, paste https://mcp.dynomapper.com/mcp, then click Connect and sign in to DYNO Mapper.

Claude Code: run claude mcp add --transport http dynomapper https://mcp.dynomapper.com/mcp, then run /mcp and choose Authenticate.

Cursor, VS Code, and other clients: add the URL as a remote MCP server and sign in when prompted. The documentation has the exact configuration for each.

Then start simple: “What DYNO Mapper projects do I have?” and build from there.

The takeaway

The first version of the DYNO Mapper MCP server showed what an AI agent could do with your audits, scores, and rankings. The hosted version removes the setup that kept most people from trying it, and the new tools let agents act on what they find: restructure the sitemap, label the pages, leave the review notes, and crawl the intranet that was never public in the first place.

One URL, one sign-in, 108 tools.

Read the documentation →