Why run AI agents in a dedicated VM?
Run AI agents on your computer, in a VM, on an old computer or in a cloud VM? See real incidents, vendor advice, RAM needs and what a VM shares.
A docs portal with Ask AI
and support tickets
Publish from GitHub, or describe your product to start.
People search, ask AI and try the API. When AI cannot answer, their message becomes a support ticket.
AI agents read the docs through MCP and llms.txt.
docs.example.com
People
- Search the docs“rate limits”
- Ask AIAnswer with links to pages
- Leave a messageOpens support ticket #12
AI agents
/mcpsearch_docs · read_page/llms.txtEvery page, with a summary/quickstart.mdEach page as Markdown
What Revdoku does
Publish from GitHub
Connect a GitHub repository, or one folder of it. Revdoku publishes its Markdown, MDX and OpenAPI files as a docs site. A Mintlify repository works as it is, with its docs.json navigation and components. No repository yet? Describe your product, and Revdoku starts the site from a docs template.
Connect the repository and choose a branch
Push to that branch, and Revdoku publishes the site again
Push another branch, and you get a preview for 24 hours
Edit a page with an AI app, and the change goes back as a commit or a pull request
acme/docsbranch: main
- mainAdd the webhooks guide✓ Published
- new-authRewrite the OAuth page✓ Preview
- revdoku/…Edits from an AI appPull request
Ask AI and support tickets
Readers ask questions with Ask AI on any page. AI answers from your docs and names the pages that it used. Each chat is kept as a ticket. If the answer does not help, the reader leaves a message, and the ticket opens for your team. Page feedback and email to the support address also become tickets. Try it on our own docs at docs.revdoku.com.
A reader asks a question with Ask AI
No answer? The reader leaves a message with an email address
The message opens the ticket for your team
The support agent drafts a reply from your docs, and a person sends it
Ask AIdocs.example.com
- Why does a webhook retry after a timeout?
- Still need help? Leave a message
- #12Webhook retries after a timeoutOpen
- AI draft · Support agent
Docs for AI agents
Each public docs site has its own MCP server at /mcp. Claude, ChatGPT, Cursor and other AI apps connect to it without sign-in. They search your docs and read each page as Markdown. The site also has llms.txt and llms-full.txt.
A developer adds
docs.example.com/mcpto an AI appThe AI app searches your docs with
search_docsIt reads the best page as Markdown with
read_pageYour analytics show which AI agents read the docs
MCP serverdocs.example.com/mcp
→ search_docs {"query": "rate limits"} ← Rate limits · /guides/rate-limits/
→ read_page {"path": "/guides/rate-limits/"} ← # Rate limits (Markdown)
- Analytics: AI agents
What every docs site has
Mintlify repositories work as they are
docs.json or mint.json navigation, MDX pages, snippets and components such as Card, Tabs, Steps and CodeGroup.
API reference and console
One page for each tag of your OpenAPI file. In the API console, people send real requests to your API.
A preview for each branch
A push to another branch makes a preview site for 24 hours. GitHub shows the link on the commit.
Ask AI on every page
AI answers from your docs and names the pages that it used. Readers rate each answer with Yes or No.
Support tickets from every channel
Ask AI chats, messages, page feedback and email arrive as numbered tickets. The support agent drafts replies from your docs.
Page feedback
Each page asks “Was this page helpful?”. The votes go to analytics, and a comment after No goes to your support tickets.
Search and three layouts
Full-text search on every site. Choose the sidebar, notebook or minimal layout in docs.json or when you publish.
Your own domain
Publish the site on your own domain. A setting hides a site from search engines.
Analytics for people and AI agents
Page views, searches, Ask AI questions and the AI agents that read your docs. A list shows what the docs miss.
Frequently Asked Questions
Does my Mintlify repository work with Revdoku?
Yes. Revdoku reads docs.json or mint.json navigation, MDX pages, snippets and Mintlify components such as Card, Tabs, Steps and CodeGroup. Components do not run code. A component that Revdoku does not know keeps its text.
Do I need a GitHub repository?
No. Describe your product, and Revdoku starts the site from a docs template. You can also upload files. If you connect a GitHub repository later, each push publishes the site.
How do AI agents read my docs?
Each public docs site has its own MCP server at /mcp, with the tools search_docs, read_page and list_pages. It needs no sign-in. The site also has llms.txt, llms-full.txt and a Markdown copy of each page. The site analytics show which AI agents read your docs. See Docs for AI agents.
What happens when I push to GitHub?
A push to the connected branch publishes the site again. A push to another branch makes a preview site for 24 hours. GitHub shows the link to the site or the preview on the commit.
Can I edit the docs in Revdoku?
Yes, with an AI app that you connect to Revdoku, or in GitHub. Your edits go back to GitHub. Choose one of three settings. Publish your edits as one commit or a pull request when you are ready, commit each edit at once, or keep the docs read-only in Revdoku.
Where do reader questions go?
To support tickets. Each Ask AI chat is kept as a ticket. If AI cannot answer, the reader leaves a message, and the ticket opens for your team. Page feedback and email to the support address also become tickets. The built-in support agent drafts a reply from your docs, and a person reads and sends it. See Support tickets.
Can I use my own domain?
Yes. Add your own domain in the site settings. Another setting hides a site from search engines.
Where does Revdoku store my data?
In the United States. Servers, the database and email run on Amazon Web Services in Northern Virginia. Files and published sites are in Cloudflare R2 in the eastern United States. TLS protects data in transit, and AES-256 encryption protects stored data.