HTML Cloud
This MCP server lets you privately share and update self-contained HTML files as end-to-end encrypted links without an account.
Share HTML privately via
share_html, passing the document inline ashtmlor as a local filepath.Encrypt locally with AES-256-GCM before upload; html.cloud stores only ciphertext and cannot read the content.
Set expiry to
7days,30days, ornever; defaults to 30 days.Add an optional public
link_nameto the share URL for readability; it is not encrypted.Receive two links: a share link for readers and a private edit link for management.
Update an existing shared page via
update_htmlusing its private edit link, replacing the page content while keeping the same share link.Use full replacements, not diffs: provide the complete updated HTML inline or via file path.
Preserve the existing link name when updating by passing the original
link_name.
Your browser encrypts the file with AES-256-GCM before anything is uploaded. The decryption key lives after the # in the share URL — browsers never send that part to the server. We store only ciphertext.
How it works
Drop an HTML file on the homepage
Your browser generates two random keys: a view key (AES-256-GCM) and an edit key
The file is encrypted in-browser and uploaded as ciphertext
You get two links:
Share link
/v/{id}#viewKey— give this to anyone you want to read the fileEdit link
/e/{id}#editKey— keep this private; it lets you replace the file, change its expiry, or delete it without changing the share link
Files self-destruct after the chosen expiry (7 days, 30 days, or never — 30 days by default)
The server stores:
The encrypted blob
The view key encrypted with the edit key (so the edit page can re-encrypt without exposing the view key)
SHA-256(editKey)for authorization — never the plaintext keys
Related MCP server: drop
Security properties
Zero-knowledge uploads — the server cannot read any uploaded content
Key-in-fragment — browsers strip the URL fragment before sending HTTP requests; the key never reaches the server
Edit authorization — only
SHA-256(editKey)is stored; the rawedit_keyis verified on every write request (replace, expiry, settings, delete)Expiry enforcement — expired documents are invisible to every route and hard-deleted daily by
model:prune10 MB upload limit enforced server-side
No accounts — the API is stateless; every state-mutating request is authorized by the edit key, and uploads are rate-limited
See html.cloud/security for the full write-up.
CLI
Share a file from the terminal — encrypted locally, same crypto module as the browser:
npx html-cloud ./file.htmlOr pipe straight from a generator:
my-report-tool | npx html-cloud -The package lives in cli/ and on npm.
See html.cloud/cli for usage and options.
MCP server
Let Claude (or any MCP client) share HTML for you. The server is published as html-cloud-mcp and lives in mcp/:
{
"mcpServers": {
"html-cloud": {
"command": "npx",
"args": ["-y", "html-cloud-mcp"]
}
}
}See html.cloud/mcp for setup instructions.
Agent skill
plugin/skills/html-cloud/SKILL.md tells an agent when to share through html.cloud and how — with the MCP server or, without it, the CLI — including updating a page at the same link instead of sharing a new copy. Copy the plugin/skills/html-cloud/ folder into your agent's skills directory (for Claude Code, ~/.claude/skills/).
Stack
Laravel 13 (PHP 8.4+)
SQLite (swap to MySQL/Postgres via
.envfor production)Vite 8 (CSS and JS build pipeline)
Web Crypto API (all crypto is browser-native; no external crypto libraries)
Setup
cp .env.example .env
php artisan key:generate
php artisan migrate
npm install && npm run build
php artisan serveRun the test suite with:
php artisan testFor production, add a cron entry to run the scheduler:
* * * * * cd /path/to/html.cloud && php artisan schedule:run >> /dev/null 2>&1Self-hosting
Clone, configure .env (APP_URL, DB_* if using MySQL/Postgres), run behind nginx or Caddy. Everything, including the encrypted blobs, lives in the database — back that up and you have the whole state.
License
Available Tools
2 toolsupdate_htmlUpdate a shared HTML pageADestructiveIdempotent
Replace the content of an HTML page that was already shared with share_html, using its private edit link. The share link stays exactly the same, so anyone who already has it sees the new version. The new HTML is encrypted locally with the same key as before; html.cloud never sees the content. Use this when the user asks to change, fix, revise, or add to a page you shared earlier in the conversation — pass the full updated HTML document, not a diff: inline as html, or as the path of an .html file as path. Use path for large pages; do not fall back to share_html if the document is too big to pass inline.
| Name | Required | Description | Default |
|---|---|---|---|
| html | No | The complete, self-contained HTML document that replaces the current one. Use path instead for large documents. | |
| path | No | Absolute path of a local .html file whose content replaces the current page, instead of passing html inline. | |
| edit_link | Yes | The private edit link returned by share_html (https://html.cloud/e/{id}#{key}). | |
| link_name | No | The link_name given to share_html for this page, if any, so the returned share link matches the one already handed out. |
TDQS
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
Annotations already declare destructiveHint=true, idempotentHint=true, and openWorldHint=true; the description adds real context beyond them: the share link is unchanged so existing recipients see the new version, content is encrypted locally with the same key, and the server never sees it. It also makes the full-replacement semantics explicit ('not a diff'), which explains why the operation is destructive yet idempotent. It stops short of documenting auth failures or error behavior, so it is not a full 5.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Five dense sentences, front-loaded with the core action and link-stability guarantee before the usage trigger and the html-vs-path mechanics. Every sentence earns its place, though the html/path advice slightly overlaps the schema's own parameter descriptions.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
For a 4-parameter mutation tool with no output schema, the description covers what changes (full document replacement), what is preserved (share link), the privacy model (local encryption), and how to supply large payloads. Nothing an agent needs to call it correctly is missing.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema coverage is 100%, so the baseline is 3, but the description adds decision guidance the schema does not: pass the full updated document rather than a diff, and prefer path over html for large pages. It also clarifies edit_link's provenance (returned by share_html) and that link_name keeps the returned share link stable.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
States a specific verb and resource ('Replace the content of an HTML page that was already shared with share_html') and explicitly ties the resource to the sibling that created it. An agent can distinguish update_html from share_html without opening either schema.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Names the explicit trigger ('when the user asks to change, fix, revise, or add to a page you shared earlier in the conversation') and adds a negative rule ('do not fall back to share_html if the document is too big'), routing the agent away from the sibling. Selection is unambiguous.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
2 tool updates
v0.1.1- Changed
share_html1 field changed- added
Input schema / properties / link_nameAdded value: +{ + "description": "A short, human-readable name added to the share link so it says what it is, e.g. \"Q3 sales report\" gives https://html.cloud/v/{id}/q3-sales-report. Unlike the page, this name is not encrypted: it is visible in the link and to the server. Use a generic description, and leave it out when the title itself is confidential (names of people, clients, deals, medical or legal matters).", + "maxLength": 80, + "type": "string" +}
- Changed
update_html1 field changed- added
Input schema / properties / link_nameAdded value: +{ + "description": "The link_name given to share_html for this page, if any, so the returned share link matches the one already handed out.", + "maxLength": 80, + "type": "string" +}
2 tool updates
v0.1.0- First observed
share_html - First observed
update_html
TDQS
Scored across 2 tools
The two tools have clearly distinct purposes: share_html creates a new encrypted share, while update_html modifies an existing share using the edit link. The descriptions explicitly guide when to use each, with update_html noting to use it instead of share_html for changes to an already shared page.
Both tool names follow a consistent verb_noun snake_case pattern: share_html and update_html. The convention is predictable and easy to understand.
With only 2 tools, the surface is quite thin for a sharing service. While each tool earns its place for the core workflow, the set feels minimal and might benefit from additional lifecycle operations.
The tools cover creation and updating of shared HTML, but there is no way to delete or revoke a share link, list existing shares, or retrieve share details. These are notable gaps in the sharing lifecycle that could cause dead ends for users wanting to manage or remove content.
Maintenance
Related MCP Connectors
Publish AI-generated HTML to a private, unguessable link, straight from Claude or ChatGPT.
Host the HTML or Markdown pages your AI generates and share each as a link with comments and access.
Instant web publishing for AI agents. POST HTML, get a live URL. No account needed.
Publish, list, update and delete shareable Markdown/HTML pages from your AI assistant.
Related MCP Servers
- AlicenseAqualityDmaintenanceMCP server for managing Netlify sites, enabling operations such as creating sites from GitHub repositories, listing, viewing details, and deleting sites.4MIT
- AlicenseNot gradedqualityDmaintenancePublishes HTML pages straight from your AI assistant to a shareable URL, then lets you manage them - update, list, search, fetch, and delete pages in a public or private workspace. Turns "share what I just made" into a single tool call from Claude, Cursor, or any MCP client.MIT
- AlicenseNot gradedqualityBmaintenanceEnables AI agents to publish HTML or Markdown to a public URL with a single HTTP POST, automatically converting Markdown to a styled webpage, with no account or setup required.1 npm1MIT

Publee MCP Serverofficial
AlicenseAqualityDmaintenanceEnables AI tools to publish HTML pages and get shareable URLs instantly, with optional account features for persistence, in-place updates, and visibility control.1MIT