hushdrop
Enables sharing artifacts (HTML, files, sites) with password protection and custom branding on your own domain, directly from any AI agent.
README
<div align="center">
Hushdrop
Share what your AI builds — on your own domain.
Open-source, self-hosted artifact sharing. Drop a file, an HTML page, or a whole site and get a branded, password-protected (zero-knowledge AES-256) link on your own domain in ~1 second — from your terminal or any AI agent. The open-source alternative to Stacktree.
Live demo → · Docs · Try in browser · vs Stacktree

</div>
Quick start
The core way to use Hushdrop is from your agent — it publishes what it builds to your own domain.
Claude Code (recommended — auto-updates via the plugin marketplace):
/plugin marketplace add maxtechera/hushdrop
/plugin install hushdrop
Cursor · Codex · Copilot · Gemini · 20+ Agent Skills hosts:
npx skills add maxtechera/hushdrop
No install — open hushdrop.dev/try, drop an HTML file, get a link. No account.
One-off from the terminal — zero install, zero setup, no account:
npx hushdrop report.html # → https://hushdrop.dev/u/xxxx (+ password, copied) — no setup, no account
Free hosted account — persistent links on your own handle (hushdrop.dev/you/<slug>):
npx hushdrop login # passwordless: GitHub or magic link
npx hushdrop report.html # → https://hushdrop.dev/you/report-a1b2 (persistent)
Your own domain — self-host on your Vercel Blob (free, MIT, your URL + brand forever):
# one-click: the "Deploy with Vercel" button above auto-provisions a Blob store — then:
npx hushdrop-install # wire your agents + CLI
hush deploy --domain share.yoursite.com # or do it from a local clone
No dashboard required. Two commands to try; one more to own.
Why this exists
Anthropic shipped artifact sharing so teams could hand each other the things they make with AI. It's great — and it's locked to their surface, their domain, their account. Stacktree does the agent-native version, but it's a closed SaaS: your content sits on their servers, custom domains and limits are paywalled, and you can't audit or self-host it.
I make things all day with agents: reports, proposals, guides, dashboards, whole little sites. I wanted
to hand someone a link that's mine — my domain, my brand, my keys — in the time it takes to type one
command, without uploading client work to someone else's server. So hush does exactly that:
terminal-native, zero-knowledge, on your own domain, usable from any agent. MIT.
Hushdrop vs. the alternatives
| Hushdrop | Stacktree | send.co | tiiny.host | |
|---|---|---|---|---|
| Your own domain | ✅ free | ❌ | ❌ | paid |
| Zero-knowledge AES-256 | ✅ | ✅ | files only | ❌ |
| Open-source / self-host | ✅ MIT | ❌ | ❌ | ❌ |
| CLI + MCP (agent-native) | ✅ | partial | ❌ | ❌ |
| Anonymous, no-account drop | ✅ | ❌ | ✅ | ✅ |
| Burn-after-read / email-gate | ✅ | partial | partial | ❌ |
| Price | free / self-host | paid | freemium | freemium |
Full write-ups: vs Stacktree · vs Send · vs ShareDuo
Features
- 🔒 Zero-knowledge — branding + AES-256 happen client-side; the server only stores ciphertext.
- 🌐 Your domain, your brand — colors, logo, social cards flow into the unlock gate + previews.
- 🤖 Agent-native — one CLI command or 9 MCP tools; any agent that runs a shell can publish.
- ⏱️ Auto-expire & burn-after-read —
--expire 7d,--burn, server-enforced cleanup. - 📧 Email-gate & revocable share links — restrict by domain, mint/revoke guest links.
- 📦 Anything static — HTML, Markdown (rendered), PDFs, images, multi-file zips → sites.
- 🚀 One-click self-host — Deploy to Vercel auto-provisions Blob; no DB to run.
What you can drop
| You run | You get |
|---|---|
hush report.html |
Branded, AES-256-locked page at yourdomain.com/<slug> (+ auto password) |
hush notes.md |
Markdown → rendered, branded HTML page |
hush report.html -p secret |
Your own password |
hush report.html --no-lock |
Branded page, no password — renders for anyone with the link |
hush report.html --expire 7d |
Auto-expire (7d/24h/2w/date); enforce with hush gc |
hush report.html --burn |
Burn-after-read — self-destructs on first view |
hush report.html --email-gate acme.com |
Only viewers with that email domain can open it |
hush deck.pdf --page |
A branded download page wrapping the file |
hush site.zip |
Multi-file static site at yourdomain.com/<slug>/ |
hush -s q3 deck.html |
Force the slug |
CLI reference
| Command | Does |
|---|---|
hush <file> |
Publish (managed / hosted / self-host, auto-detected) |
hush login / whoami / logout |
Passwordless hosted account (GitHub or magic link) |
hush list / rm <slug> / gc |
List, delete, garbage-collect drops |
hush share <slug> [--revoke] |
Mint / revoke a revocable guest link |
hush claim <url> |
Move an anonymous /u/ drop into your account |
hush set-expiry / set-email-gate / set-password / set-feedback |
Manage a hosted drop |
hush init / setup / deploy |
Configure + self-host on your own domain |
hush migrate |
Re-home hosted drops onto your own Blob |
Use it from any AI agent
npx hushdrop-install registers the drops MCP server into your detected agents (Claude Code, Codex,
Cursor, Windsurf, OpenCode, Amp), puts hush on your PATH, and prints config for GUI clients.
claude mcp add hushdrop -- npx -y hushdrop-mcp # Claude Code
codex mcp add hushdrop -- npx -y hushdrop-mcp # Codex
{ "mcpServers": { "hushdrop": { "command": "npx", "args": ["-y", "hushdrop-mcp"] } } }
MCP tools (9 — vs Stacktree's 7)
| Tool | Purpose |
|---|---|
publish_html |
Publish raw HTML → branded, password-protected link |
publish_file |
Publish a local file (optionally a branded download page) |
update_site |
Replace a drop's content in place (same URL) |
list_sites / delete_site |
List / delete drops |
set_password / set_expiry / set_email_gate / set_feedback |
Manage a hosted drop |
How it works
- Read your file and detect its type.
- Brand (HTML) — inject favicon, OG/Twitter card, and a subtle corner badge before
</body>. - Encrypt (if locking) — StatiCrypt (AES-256, client-side) behind your branded unlock gate. The badge is baked in before encryption, so it survives.
- Upload to Vercel Blob under a clean key.
- Serve —
yourdomain.com/<slug>proxies the blob viamiddleware.js, rewriting headers so encrypted HTML decrypts + renders (not downloads) and CSP doesn't block the unlock script. Drops arenoindex, nofollow, noai. - Report — URL (+ password) printed and copied to your clipboard.
Serving is a dumb transparent proxy; all branding + encryption happen client-side at upload. The server only ever stores ciphertext.
Self-host
The Deploy button clones the repo and auto-provisions a Blob store (it injects BLOB_READ_WRITE_TOKEN;
the app derives its store from the token, so it just works). Then npx hushdrop-install + hush init --domain share.yoursite.com to point your CLI + agents at your instance. Self-host is token-only, free, unlimited
— no account needed. Full walkthrough in skill/SETUP.md.
Configuration
| File | Holds | Committed? |
|---|---|---|
skill/brand/brand.json |
Presentation — name, colors, owner, social links. Edit to rebrand. | yes |
~/.hushdrop/config.json |
Infra — domain, blob host, Vercel project. Written by hush init. |
no |
~/.hushdrop/.env |
Your BLOB_READ_WRITE_TOKEN. |
no |
FAQ
Is it really zero-knowledge? Yes — for locked drops, AES-256 runs in your browser before upload; the server stores only ciphertext.
Managed vs. hosted vs. self-host? Managed = anonymous, 24h, no account. Hosted = free account, persistent links on hushdrop.dev/you/…. Self-host = your own domain + Blob.
Does it need Vercel? Only to self-host. The managed + hosted tiers need nothing but npx.
How is this different from Stacktree? Same agent-native idea, but open-source, self-hostable, your own domain free, and anonymous (no account). See the comparison.
Can agents use it without MCP? Yes — it's a single CLI; any agent that runs a shell command can publish.
Is it free? Yes. MIT. Self-host costs only your own (usually pennies) Vercel Blob usage.
Security
- Locked HTML is genuinely AES-256 encrypted in the browser — use long passwords; strong against casual access, not a vault.
- Raw files are protected by an unguessable slug; use
--page -p <password>for a gated download. - Passwords are stored in
~/.hushdrop/manifest.jsonon your machine only — never uploaded.
See SECURITY.md to report a vulnerability.
Contributing
PRs welcome — see CONTRIBUTING.md. It's a few hundred lines of readable Node + HTML, no build step.
<div align="center">
MIT · Built by Max Techera · GitHub · Instagram · hushdrop.dev
</div>
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.