Postiz
Open-source social media scheduling with a public API, workflow integrations and an agent CLI
Postiz is an open-source social media scheduling tool — the project positions itself against Buffer and Hypefury — that plans, queues and publishes posts to X, LinkedIn, Reddit, Instagram, TikTok, YouTube, Facebook, Threads, Bluesky, Mastodon, Discord and others from one calendar. It is AGPL-3.0 TypeScript, self-hostable as a Next.js and NestJS stack on PostgreSQL, and its README states there is currently no feature difference between the hosted and self-hosted versions. It had 34,927 stars on GitHub as of 21 August 2026. What sets it apart from a dashboard-first scheduler is the surface area around the queue: a documented public API, a Node SDK, n8n and Make.com integrations and a separate agent CLI, so something other than a person can fill it.
What Postiz actually is
Postiz (github.com/gitroomhq/postiz-app) is a scheduling and publishing layer for social accounts. You connect each account through its own platform OAuth flow, compose once, and a calendar decides when each network receives the post. The repository has been going since July 2023 and is genuinely worked on rather than parked: releases run to v2.23.0 on 4 August 2026, and the commits in the days before we captured this page are feature work — MCP endpoints, an exposed OAuth controller, security and framework bumps — not README housekeeping.
- Reach Fourteen networks are listed in the README, among them X, LinkedIn, Reddit, Instagram, TikTok, YouTube, Facebook, Pinterest, Threads, Bluesky, Mastodon, Discord and Slack.
- Self-hosting A pnpm monorepo — Next.js front end, NestJS API, Prisma against PostgreSQL, Temporal driving the schedule — with a documented quick start. The README says the hosted service and the self-hosted build are currently the same product.
- Automation surface A public API, a Node SDK on npm, an n8n community node, a Make.com integration and a separate agent CLI repository.
- Team workflow Several members on one workspace, comments on posts before they go out, and analytics on what was published.
- Licence AGPL-3.0. If you modify it and offer it to others over a network, the copyleft terms reach your changes — read that before folding it into a product you sell.
How it compares with the other open-source scheduler we have covered
We have already written up BrightBean Studio, which does the same job: plan, schedule, approve and publish across many networks, self-hosted, using your own platform credentials. If you are choosing between them, the feature lists will not settle it, because both cover the obvious ground. The honest difference is who is expected to drive.
BrightBean is a dashboard a person sits in front of. Postiz is far larger by adoption, carries more integrations, and puts a documented API and an agent CLI in front of the same calendar, which makes it the easier one to wire into automation you already run. That cuts both ways. The moment a queue can be filled by a script, the interesting question stops being the calendar and becomes who checked what went into it.
The part that gets hard when a human is not typing
We build in this area ourselves, and the lesson that cost us most had nothing to do with scheduling. Publishing later than you approve opens a window, and inside that window the content can change. A queue that treats approval as a flag on a row, rather than as consent to specific words, will eventually publish something nobody agreed to — and it will look like it worked.
- Approval is bound to the bytes, not the row. A digest of the post body is stamped when a person approves it and re-checked at the moment the slot fires; if the two no longer match, the slot reverts to draft instead of publishing.
- An approval carries the version the reviewer was looking at. When an edit lands mid-review, the approval comes back as skipped rather than being quietly applied to text nobody read.
- Our character counter for X warns instead of blocking, because the real limit depends on the account's subscription tier and no scheduler can know which one you hold. Refusing a post that would have been fine is a worse failure than shipping a warning.
What a queue does not decide
A full calendar is not an audience. Scheduling removes the friction of posting consistently, which is necessary and nowhere near sufficient: it does not decide what is worth saying, which network deserves the effort, or whether anyone is reading. Those are the questions in how often should a founder post on X.
It also does not tell you whether a post truly went out. A platform API can accept a call and still decline the post, so the answer that counts is the one you read back from the network rather than the one your scheduler recorded. That failure, and how to gate around it, is covered in should I let AI post to social media for me.
FAQ
- Is Postiz free?
- The source is free under AGPL-3.0 and you can self-host it; what you pay for is the server, the database and the time you spend keeping it running. The project also sells a hosted service, and its README states there is currently no feature difference between the hosted and self-hosted versions.
- How is it different from Buffer or Hootsuite?
- Ownership and extensibility. Self-hosting puts the schedule and the content on your own infrastructure rather than a vendor's, with no per-seat or per-channel pricing, and the public API, SDK and agent CLI let you drive the queue from your own systems. The trade is that you operate it: a database, a scheduler and a set of platform app registrations become yours to keep working.
- Can an AI agent post through it?
- Mechanically, yes — it exposes a public API and ships a separate agent CLI. Whether it should is a policy question rather than a technical one. Our own rule is that anything a stranger will see waits for a person to approve the exact text, while reversible actions do not; the reasoning is in should I let AI post to social media for me.
- What does it take to run?
- A PostgreSQL database and a Node runtime for the Next.js and NestJS services, with Temporal handling the schedule — most simply through the documented Docker quick start. Budget more time for the platform side than the server side: every network you connect needs its own developer app and, on several of them, a review process. Those are the steps that stall.
You built it. Now grow it.
AgentCeres is a managed AI marketing team — specialists draft the SEO, social, and outreach that fill your links, you approve what ships. 14-day free trial, from $39/month.