Copilot writes the docs,
Revdoku publishes them

Revdoku publishes a docs site from a GitHub repository: Markdown and MDX pages, docs.json navigation and an OpenAPI file. Copilot changes these files in a branch, and you review them in a pull request.

1. Create a Revdoku account and connect the GitHub repository of your docs.
2. Put the rules for your docs in AGENTS.md. Copilot uses the nearest AGENTS.md file in the directory tree. The example rules work there too.
3. Assign an issue to Copilot, or ask Copilot in VS Code to change the docs.
4. Push the branch. The cloud agent pushes its own branch when it opens the pull request. Revdoku builds a preview site of the branch for 24 hours.
5. In the pull request, the Revdoku preview status links to the preview site.
6. Merge the pull request into the connected branch. Revdoku publishes the site. On GitHub, the new commit gets a Revdoku status that links to the site.

Revdoku makes branch previews only while the main site is live. A pull request from a fork gets no preview, because the fork pushes to another repository.

GitHub Copilot in VS Code 1.99 or later can call MCP servers (outside services that give the agent extra tools). Each public Revdoku docs site has its own read-only MCP server at /mcp. It needs no sign-in.

1. Open the Command Palette and run MCP: Add Server.
2. Choose HTTP (HTTP or Server-Sent Events). Enter the address of the docs site with /mcp at the end, for example https://docs.example.com/mcp.
3. Enter the server ID acme-docs.
4. For a personal setup, select Global, then Copilot Global.
5. Open Copilot Chat and select Agent. Ask: “Use the acme-docs tools. Which pages describe webhooks?”
6. Expand the tool call in the chat. Make sure that it came from the acme-docs server.

You can also write the configuration by hand. For yourself, add this entry to ~/.copilot/mcp-config.json. For your team, add it to .mcp.json in the project root:

{
  "mcpServers": {
    "acme-docs": {
      "type": "http",
      "url": "https://docs.example.com/mcp"
    }
  }
}

The older .vscode/mcp.json format puts the same entry under "servers" instead of "mcpServers". Copilot gets three tools: search_docs, read_page and list_pages.

The docs server only reads a site. To publish your sites and read support tickets from Copilot, add the Revdoku server for your account. It signs you in through the browser, so no API key goes into VS Code.

1. Run MCP: Add Server again. Choose HTTP (HTTP or Server-Sent Events) and enter https://mcp.revdoku.com.
2. Enter the server ID revdoku. Select Global, then Copilot Global.
3. When VS Code asks you to sign in, finish the Revdoku sign-in in your browser. Select the buckets and an access level. To publish and preview sites, Copilot needs Full access.
4. In Agent mode, ask: “Use the Revdoku MCP tools to list my buckets. Do not change files.”

A bucket holds the files of one docs site, the site and its support tickets. The configuration file holds no key. Every person who uses the file signs in to Revdoku separately.

If a bucket gets its files from GitHub, change the docs in the repository. The Edits in Revdoku setting of the bucket decides what happens to an edit that Copilot makes in Revdoku. The edit waits for a person, becomes a commit, or Revdoku refuses it (Read-only).

On Copilot Business and Copilot Enterprise, MCP works only after an admin turns on the MCP servers in Copilot policy. The policy is off by default. It does not apply to Copilot Free, Pro, Pro+ or Max. An organization can also limit which MCP servers are available.

The Copilot cloud agent works on GitHub.com in its own cloud environment and opens pull requests. GitHub says that this agent does not support remote MCP servers that use OAuth, so it cannot use the Revdoku account server. The docs server needs no sign-in, so the cloud agent can use it.

1. On GitHub.com, open the repository and click Settings.
2. In the sidebar, under Code, planning, and automation, click Copilot, then MCP servers.
3. Paste this configuration and click Save MCP configuration:

   {
     "mcpServers": {
       "acme-docs": {
         "type": "http",
         "url": "https://docs.example.com/mcp",
         "tools": ["search_docs", "read_page", "list_pages"]
       }
     }
   }

4. Assign an issue to Copilot. In the session log, expand the Start MCP Servers step. Make sure that the three docs tools are in it.

The cloud agent uses MCP tools without asking for approval, so GitHub recommends a list of read-only tools. The three docs tools only read public pages. The agent can then compare its code with the published docs before it opens the pull request.

• API reference. Copilot changes an endpoint and the OpenAPI file in the same pull request. Revdoku makes one reference page for each OpenAPI tag and an API console.
• Code samples. Copilot reads the published pages through the docs server and fixes samples that no longer match the code.
• Support questions. In VS Code, with your account connected, Copilot reads the support tickets. It drafts a page for each question that the docs do not answer.

When a tool call needs approval, expand it and read the inputs. Select Allow only for an action that you understand. A prompt for the first task:

Use the acme-docs MCP tools. Read the published page about orders.
Compare it with the order endpoints in src/api/ and docs/openapi.json.
Fix docs/openapi.json and the pages that do not match the code.
Push the changes to a new branch and open a pull request.

GitHub: MCP in Copilot Chat · GitHub: MCP for the cloud agent · VS Code: add MCP servers · Docs for AI agents · Revdoku MCP guide