PostSieteWhat "self-hosted social media scheduler" actually costs to run
Most social media tools bill per channel. Seven channels means seven line items, and the more places you show up, the more you pay — which is exactly backwards for a one-person marketing team whose whole problem is having to be in seven places at once.
The alternative people reach for is self-hosting, and the usual objection is that it is false economy: you save the subscription and pay it back in server bills and weekends. That is often true. It is worth being specific about when it is not.
The three costs, separately
Self-hosting a drafting-and-scheduling stack has exactly three costs, and conflating them is how people end up surprised.
The server. A small VPS. This kind of workload is not CPU-bound and not memory-hungry — it is a web process that spends most of its life idle, waiting for someone to type an idea. The cheapest tier at any of the usual providers handles it. Call it the price of a coffee a month.
The AI. This is the one that actually varies, and it is the one nobody quotes you. Generating a set of platform-native drafts from one idea costs somewhere around 2,800 output tokens plus roughly 1,200 in. At current API rates that is fractions of a cent per set. Thirty sets a month — a post a day — is still comfortably under a dollar.
That number is worth sitting with, because the AI tier of every hosted tool in this category is priced as an upsell, at a multiple of what the underlying tokens cost. When you bring your own key, you pay the token price. The gap between those two numbers is the entire business model of a lot of AI SaaS.
Your time. The real cost, and the one that decides it. A stack that needs a build step, a database migration, a queue worker and a reverse proxy will take a weekend to stand up and an hour a month forever. A single Node process with no build step will not.
Where self-hosting is genuinely wrong
It is worth saying plainly, because the honest version of this argument has a limit.
If you have never deployed anything, the weekend you spend learning is worth more than the subscription you saved. If you need publishing to never fail unattended, a hosted service with someone on call is a real product and you are buying the on-call, not the software. And if the tool is not the bottleneck in your marketing — if the bottleneck is that you do not know what to say — then no amount of infrastructure ownership fixes that.
Where it is right
You already run servers. You want your content, your API keys and your connected accounts to live somewhere you control. You would rather pay token prices than AI-tier prices. Or you simply object, on principle, to a per-channel bill for software that does not cost more to run per channel.
That last one is not irrational. Writing seven native drafts is the expensive part of this problem, and that cost does not multiply per channel — so a price that does is a pricing decision, not a cost pass-through.
What we do about it
PostSiete is one Node process. No build step, no bundler, and one runtime dependency that is loaded lazily and only when you point it at Postgres. Put it on a small VPS with your own model key and the software cost is zero, permanently. The hosted version is twelve dollars flat for all seven platforms, because the same argument applies to us: it does not cost us more per channel either.
Same code, both ways. That is the point — if the self-hosted version were a crippled build, the claim would be marketing rather than a fact.