Swolvy

Swolvy MCP server and skill: publish HTML from Claude, ChatGPT, Gemini and coding agents

Swolvy turns an HTML file into a shareable link. Drop a file on swolvy.com, or let your AI assistant (Claude, ChatGPT, Gemini) or coding agent publish it for you, and get a short link with optional password protection. No signup is needed. A page can be replaced later without changing its link, and Swolvy Pro adds pages that never expire, files up to 100 MB and your own domain.

This repository is the public home for the two ways AI assistants publish through Swolvy:

• the remote MCP server at https://swolvy.com/api/mcp, for Claude, ChatGPT, Gemini, Claude Code, Cursor and any other MCP client
• the `swolvy` skill for Claude Code, in skills/swolvy, built on the HTTP API

Swolvy itself is not in this repository. server.json describes the server in the MCP Registry format.

Add the MCP server

One address, OAuth sign-in, no API key. The server works without a Swolvy account: anonymous uploads return a link and an edit token. During sign-in you can also link the connector to a Swolvy account, so pages land in your dashboard and the account tools below are added.

https://swolvy.com/api/mcp

The transport is Streamable HTTP with OAuth 2.1 (dynamic client registration and PKCE).

Claude (claude.ai and desktop)

1. Open Settings, then Connectors.
2. Choose Add custom connector.
3. Paste https://swolvy.com/api/mcp and confirm.
4. Sign in when asked: continue without an account, or link your Swolvy account.

Claude Code

claude mcp add --transport http swolvy https://swolvy.com/api/mcp

Run /mcp inside Claude Code to complete the OAuth sign-in. Add --scope user to make the server available in every project.

ChatGPT

On chatgpt.com: Plugins, then +, then Add custom MCP server. Paste the address, choose OAuth as the authentication, then install the plugin. The desktop app's Add MCP server form (Streamable HTTP) works too, but without account linking.

Cursor

Add the server to .cursor/mcp.json in your project, or to ~/.cursor/mcp.json for every project:

{

"mcpServers": { "swolvy": { "url": "https://swolvy.com/api/mcp" } } }

Gemini

In the Gemini app: Settings & help, then Connected Apps, then Add a custom app, and paste the address.

For Gemini CLI, add the server to ~/.gemini/settings.json:

{

"mcpServers": { "swolvy": { "httpUrl": "https://swolvy.com/api/mcp" } } }

Any other MCP client

Add a remote server (Streamable HTTP) with the same address. Clients that support OAuth sign in through the browser; the server advertises its metadata at https://swolvy.com/.well-known/oauth-protected-resource/api/mcp.

Tools

Every connector gets the four publishing tools. A connector linked to a Swolvy account also gets the account tools. The team tools and the workspace parameter appear when the account belongs to a Swolvy Team.

Publishing tools (always available)

The preferred large-file flow is: start_swolvy_upload, then one PUT of the whole file to upload_url (for example curl -X PUT -H 'Content-Type: text/html' --data-binary @page.html ), then finish_swolvy_upload.

Password handling: pass password when the user named one, or generate_password: true when they only asked for protection. Swolvy then picks a memorable password (like amber-river-42) and returns it as password; the assistant tells the user. Pages can be replaced up to 10 times with replace_id and replace_edit_token; the new version keeps the current password unless a new password or remove_password: true is given.

Account tools (connector linked to a Swolvy account)

Team tools (Swolvy Teams)

Upload with the HTTP API

For scripts, CI jobs and coding agents that would rather make one HTTP request than add an MCP server.

Request

POST https://swolvy.com/api/v1/upload with a JSON body, or a multipart/form-data body with the same field names (plus file for the HTML file itself).

Authentication is optional. Without it the upload is anonymous. With Authorization: Bearer swk_... (an API key from the dashboard, under AI & API) the page is saved to your account.

curl

curl -X POST https://swolvy.com/api/v1/upload \

-H "Content-Type: application/json" \ -d '{"html":"<!DOCTYPE html> Hello ","filename":"Hello.html"}'

Or send a file from disk:

curl -X POST https://swolvy.com/api/v1/upload \

-F "file=@index.html" -F "filename=index.html"

Response

{

"url": "https://swolvy.com/abc123", "id": "abc123", "expires_at": "2026-11-21T10:00:00.000Z", "edit_token": "...", "password_protected": false }

Other fields appear when they apply: password (when generate_password was used), never_expires: true (Pro), replaced: true, path, title, description, custom_url (the page's link on a custom domain), workspace, and viewer_warning (visitors see a short notice before a page that looks like a sign-in form). Errors are JSON with one field: {"error": "..."}.

Keep the edit_token: it is the only way to replace an anonymous page later.

Python

import requests

r = requests.post( "https://swolvy.com/api/v1/upload", json={"html": open("index.html").read(), "filename": "index.html"}, ) r.raise_for_status() page = r.json() print(page["url"], page["edit_token"])

Node

import { readFile } from "node:fs/promises";

const res = await fetch("https://swolvy.com/api/v1/upload", { method: "POST", headers: { "Content-Type": "application/json" }, body: JSON.stringify({ html: await readFile("index.html", "utf8"), filename: "index.html" }), }); const page = await res.json(); if (!res.ok) throw new Error(page.error); console.log(page.url, page.edit_token);

Replace a page

Send the new HTML with the id and edit token from the first response. The link does not change and the 30-day clock restarts.

curl -X POST https://swolvy.com/api/v1/upload \

-H "Content-Type: application/json" \ -d '{"html":"<!DOCTYPE html> Hello again ","filename":"Hello.html","replace_id":"abc123","replace_edit_token":"..."}'

Large files (upload sessions)

One request carries up to 2 MB. Pro accounts can send pages up to 100 MB through a session; the response's max_bytes says what the current plan allows.

1. POST https://swolvy.com/api/v1/upload/session with {"filename": "report.html"} and the same optional fields as above (password, generate_password, replace_id, replace_edit_token, remove_password, path, title, description, workspace). Send Authorization: Bearer swk_... for a Pro account. Response: {"upload_id", "upload_url", "finish_url", "max_bytes"}.
2. PUT the raw HTML to upload_url: curl -X PUT -H "Content-Type: text/html" --data-binary @report.html .
3. POST to finish_url (/api/v1/upload/session/{id}/finish). The response has the same fields as /api/v1/upload.

Rate limits

Uploads are limited to 10 per hour per IP address (HTTP 429), with a global cap across all callers. Bodies over the size limit return 413.

Claude Code skill

The swolvy skill teaches Claude Code the API call above so that "publish this page" or "give me a link to this HTML" works in one step.

Install it by copying the skill folder into your Claude Code skills directory:

git clone https://github.com/enahmedsamy/swolvy-mcp.git

mkdir -p ~/.claude/skills cp -r swolvy-mcp/skills/swolvy ~/.claude/skills/swolvy

Or add it to one project only, under .claude/skills/swolvy/. Then ask Claude Code to publish an HTML file, or run /swolvy. The skill file is skills/swolvy/SKILL.md.

If you would rather use tool calls than curl, add the MCP server to Claude Code instead (see above). Both give the same link.

Limits and plans

| Replace on