Teams using Claude Code, Codex or Cursor
Let your coding agent publish and revise client work
Connect your agent to QuickHost once, over MCP or the qh CLI. Then ask it to publish the deliverable it just built, protect it with a password and give you the link. When your client comments, ask it to read the thread, make the changes and republish to the same address.
By the QuickHost team. Checked against the product on .
The situation
Where does this usually go wrong?
Your agent can build the deliverable quickly, and then the handoff is manual: you export the files, upload them somewhere, set the access and copy the client’s feedback back into the chat. Each revision repeats those steps.
- The agent finishes, and you do the uploading and the access settings by hand.
- Client feedback arrives by email, and you paste it into the chat for the agent.
- Each revision is a new file, and on some hosts a new link to send.
Step by step
How do you do it with QuickHost?
Connect once, then ask in plain language. The agent uses the same operations you’d use in the dashboard.
- 1
Connect once
Follow the MCP setup page for Claude Code, Claude Desktop, ChatGPT, Cursor or Codex: one command or one pasted URL, then a sign-in in your browser.
- 2
Ask it to publish
Tell it to publish ./dist as acme-app, password-protected. It comes back with the live URL and the password.
- 3
Ask for named links
It can create a share link for each person at the client, each with its own expiry date if you want one.
- 4
Hand it the feedback
When the client comments, ask the agent to read the comments. It gets each one with its replies and the text it was left on.
- 5
Let it fix and republish
It edits the files, republishes to the same address, and replies to or resolves the comments it addressed.
What you get
What do you get from it?
Publishing, access and feedback go through the same agent that built the work, so a revision round needs one request from you.
One request per revision round
Reading the comments, editing, republishing and resolving happen in the same session.
Comments the agent can act on
Each comment comes with its anchor, so the agent knows which sentence or page it’s about.
A link that survives revisions
Republishing keeps the address and every earlier version.
Access you can take back
Connected apps are listed in your settings with a Disconnect button, and API tokens can be labelled and revoked.
Other options
What else could you use for this?
These tools come up for the same job. Each comparison is dated and sourced from the other product's own pages, and says when it's the better pick.
- QuickHost vs Claude artifacts
Claude Code can publish an artifact itself, as one file with no password, expiry date or open tracking.
- QuickHost vs display.dev
display.dev has a complete agent comment loop too, aimed at work your own company reads.
Questions
What else do people ask about this?
The questions that come up when someone is setting this up for the first time, answered plainly.
Which agents can publish to QuickHost?
Any MCP client, including Claude Code, Claude Desktop, ChatGPT, Cursor and Codex, plus anything that can run the qh CLI or call the REST API.
Can the agent read my client’s comments?
Yes. Over MCP or with qh comments --json it gets each comment, its replies and the text or page it’s attached to, and it can reply to and resolve them.
Does the agent need my password?
No. It signs in through OAuth in your browser, or uses an API token you create and can revoke at any time.
Can the agent publish things I haven’t seen?
It publishes only when you ask it to, under your workspace, and every link traces back to your account. You can review the project in the dashboard before you send it.