Skip to main content
pendpost is local-first: the scheduler is an in-process tick on your machine. YouTube posts survive your computer being off because the platform itself fires them (YouTube publishAt), and Mastodon, WordPress, and Ghost posts survive the same way via their own native scheduling. pendpost publishes Instagram, Facebook, LinkedIn, X, Telegram, Discord, and Nostr itself at the scheduled time, so pendpost must be running then (or the optional 24/7 cloud service must cover them). To close that gap for free, run pendpost on a host that never sleeps. The one rule: never put an unauthenticated publish API on the public internet.

The honest power-off split

So you only need always-on hosting if you schedule Instagram, Facebook, LinkedIn, X, Telegram, Discord, or Nostr. The safest pattern keeps the pendpost server exactly as it ships, bound to loopback with no auth, and puts access control in front of it with a tunnel. Nothing about the server changes, and the publish API is never exposed.
  • Tailscale: run pendpost on a small always-on host (a cheap VPS, a Raspberry Pi, or a NAS) and reach it over your tailnet. Optionally use Tailscale Serve to terminate TLS on the tailnet only.
  • Cloudflare Tunnel + Access: run cloudflared next to pendpost pointing at http://127.0.0.1:8090, and put a Cloudflare Access policy (email or SSO) in front. Only authenticated people reach it; the server stays loopback-only.
Because the server stays loopback-bound, its default posture is unchanged: no token, no new surface.

Containers and one-click hosts

The repo ships a Dockerfile and docker-compose.yml. Compose keeps host exposure on 127.0.0.1 while the container binds 0.0.0.0 internally, so a local always-on box stays loopback-only:
For managed always-on platforms, copy a template from deploy/:
  • deploy/fly.toml for Fly.io (with a persistent volume and a health check)
  • deploy/railway.json for Railway
  • deploy/render.yaml for Render
These platforms give you a public URL, so you must gate access. Use the optional in-server auth gate below, or front the service with Cloudflare Access.

Optional: the in-server auth gate

When you cannot put a proxy in front, pendpost can require a bearer token itself. The gate is off by default and activates ONLY when both are true:
  1. PENDPOST_HOST is a non-loopback address (for example 0.0.0.0 or a Tailscale IP), and
  2. PENDPOST_AUTH_TOKEN is set.
With the loopback default, the gate is always off and the local posture is byte-for-byte unchanged. When it is on, every request must carry Authorization: Bearer <PENDPOST_AUTH_TOKEN>, except the readiness probe GET /api/health so platform health checks keep working.
Drive it over the API or MCP with the token:
The dashboard SPA cannot send a bearer header on its own, so when the gate is on, use the dashboard over the loopback tunnel and reserve the public URL for the API and MCP. The token is the security boundary: keep it secret, set it as a platform secret (never in a committed file), and rotate it if it leaks.

What this does not change

The approval gate still holds everywhere: an always-on host only fires posts that were already approved. Nothing publishes until it’s approved; by default that’s you, and auto-approve is owner-only and revocable. Hosting changes where pendpost runs, never who is allowed to publish.