publish-artifacts-mcp
Enables publishing, updating, and sharing HTML artifacts with strict security isolation (origin separation, CSP, API keys) via MCP tools.
README
publish-artifacts-mcp
Publish self-contained HTML artifacts safely, controlled by API key, and manage them over MCP (stdio + Streamable HTTP). Minimal admin UI included.
The hard part here is not "serve HTML" — it's hosting untrusted HTML without letting it steal credentials or exfiltrate data. The design centers on origin isolation + strict CSP + scoped API keys.
Stack
TypeScript · Node 22 · Hono · @modelcontextprotocol/sdk
· SQLite (better-sqlite3, WAL) · zod. Deploy via Docker.
How it works
Two planes share one process and one SQLite file, but are routed by Host header:
| Plane | Origin (example) | Serves | Auth |
|---|---|---|---|
| Control | api.example.com |
REST API, admin UI (/admin), MCP-over-HTTP (/mcp) |
API key (Bearer) |
| Content | content.example.com |
published artifacts at /a/:slug |
public, no cookies |
They must be different origins — the server refuses to boot otherwise (set
ALLOW_SAME_ORIGIN=1 only for local dev). This guarantees untrusted artifact
HTML can never reach the API/admin surface.
Safety layers
- Origin isolation (above) + host-based routing.
- Strict CSP on every served artifact:
default-src 'none',connect-src 'none',form-action 'none',frame-ancestors 'none',base-uri 'none'— blocks network/form exfiltration and framing while still allowing the artifact's own inline script/style to render.relaxedmode (opt-in per artifact) permits https network access. - Hardening headers:
X-Content-Type-Options,Referrer-Policy,X-Frame-Options,Permissions-Policy,Cross-Origin-Resource-Policy. No cookies are ever set on the content origin. - Unguessable slugs (~132 bits), optional expiry, and rotate to invalidate a shared link.
- API keys: high-entropy, SHA-256-hashed at rest, prefix-indexed, revocable, and owner-scoped (a key sees/mutates only its own artifacts).
- Limits: max HTML size + per-key publish rate limiting. Optional
sanitizetier strips scripts via DOMPurify.
Accepted trade-off (raw-path model): all artifacts share one content origin, so they can read each other's
localStorage. That's why no credentials live on the content origin. Upgrading to per-artifact subdomains (<slug>.content-domain) later requires only routing + wildcard DNS changes.
Quick start (local dev)
npm install
cp .env.example .env # defaults use content.localhost for isolation
npm run build
# Bootstrap an API key (printed once):
npm run create-key -- "my laptop"
# Run the server:
npm start
# admin UI: http://localhost:8787/admin (paste the key when prompted)
content.localhost resolves to 127.0.0.1 on most systems. If yours differs, add
127.0.0.1 content.localhost to /etc/hosts, or set ALLOW_SAME_ORIGIN=1 for
dev only.
MCP tools
publish_artifact, update_artifact, delete_artifact, get_artifact,
list_artifacts, get_share_link. All owner-scoped by the calling key.
stdio (Claude Code / Desktop) — .mcp.json
{
"mcpServers": {
"artifacts": {
"command": "node",
"args": ["/abs/path/dist/mcp/stdio.js"],
"env": {
"ARTIFACTS_API_KEY": "ak_...",
"CONTENT_BASE_URL": "https://content.example.com",
"CONTROL_BASE_URL": "https://api.example.com",
"DB_PATH": "/data/artifacts.db"
}
}
}
}
Remote (Streamable HTTP)
POST https://api.example.com/mcp with header Authorization: Bearer ak_....
Deploy (Docker)
Point two hostnames you control at the container (reverse proxy / DNS), set them
as CONTROL_BASE_URL / CONTENT_BASE_URL in docker-compose.yml, then:
docker compose up -d --build
docker compose exec artifacts node dist/cli/createKey.js "prod key"
SQLite persists in the artifacts-data volume (/data).
Tests
npm test # unit + MCP in-memory + HTTP host-isolation integration
The only
npm auditfindings are in the dev-onlyvitest/vite/esbuildchain (not shipped to production). Left unpinned to avoid a breaking vitest v4 bump; revisit when upgrading the test tooling.
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.