Run your socials from the command line
or hand it to an agent

Overview

Creator OS is the operating system for social media: one account connects Instagram, TikTok, YouTube, X, Facebook, LinkedIn, and Threads, and everything you can do in the dashboard (publishing, scheduling, analytics) is also available through the Creator OS CLI. That makes the whole platform agent-ready: any AI agent can run an account end to end with one scoped API key.

These docs cover the two ways to set that up, the Kairos agent, and, workflow by workflow, the endpoint surfaces your agent works with: comment funnels, replies, moderation, DMs, multi-platform posting, scheduling, and analytics. No REST calls to learn; the workflows below are the interface.

Quickstart

First, get your API key. Create a Creator OS account, connect your socials, and grab the key from Settings → API keys (one key per workspace; agency plans get one per client). Keys look like sk_ + 64 hex characters.

From there, there are two ways to set up. Most people should run the Creator OS agent: the CLI ships inside its codebase, so there is nothing separate to install.

Either way, on that first run the agent checks your connected accounts, pulls follower stats and analytics, and reports back with everything it can do. From then on you talk to it in plain language: “schedule this week’s clips,” “post this everywhere,” “how did last week perform?”

Kairos, the Creator OS agent

Kairos (the Kai agent) is Creator OS’s own agent harness: a ready-to-run agent with the CLI built into its codebase. Instead of wiring the toolkit into your own AI coder, you run Kairos and get an agent that already knows it. It onboards you with a few setup questions, writes a brand file it re-reads before every run, and manages your accounts end to end.

Kairos ships as a fork-and-run starter repo: github.com/kevinbadi/kairos (open source, MIT). Fork or clone it, npm install(bootstraps the CLI and agent surfaces), set your Anthropic API key (the agent's brain), and start. Your Creator OS workspace key is collected during the onboarding interview, masked as you type:

Onboarding: Kairos builds your brand file

On first run Kairos interviews you and saves the answers to a brand file (brand.md) that anchors everything it writes and replies to:

What Kairos can do

This is the exact Creator OS API surface Kairos ships with. Every other endpoint is gated off from the agent, so it can’t touch anything outside this list.

Comments → DM

Works on: Instagram, Facebook

The funnel workflow: someone comments a keyword on your post (“GUIDE”, “LINK”, “PRICE”) and they automatically get a DM with your link. It’s the highest-converting thing you can run on Instagram, because the platform rewards comment volume and the DM lands while the person is still looking at your content.

A comment-to-DM automation has a trigger, a match rule, and a payload:

Reply to comments

Works on: Instagram, Facebook, X, Threads, YouTube, LinkedIn (every platform except TikTok)

Every comment across every connected account flows into one inbox. List them all, pull one post’s thread, and reply, from the CLI or the Creator OS app’s Inbox tab.

There’s also a private-reply move most people don’t know exists: answer a comment in the commenter’s DMs. Right for “what’s the price?” questions you’d rather not negotiate in public.

For hands-off replies, AI auto-replies run server-side: configure the voice prompt and your links in the Creator OS app and incoming comments get answered around the clock, no CLI or machine involved. Kairos does the same thing webhook-driven, with judgment: it drafts on-brand replies and escalates sensitive threads to you.

Like comments

Works on: Facebook, X

The lightest-touch engagement there is: liking a comment tells the commenter they were seen without you writing anything, and comment engagement feeds the platforms’ ranking of your post. A like-sweep after posting (every positive comment in the first hour) is a cheap, real boost.

Platform note: Instagram removed comment-liking from its API back in 2018, and Threads, YouTube, and LinkedIn don’t offer it either. That’s why this one is Facebook and X only; on the other platforms, a short reply is the equivalent gesture.

Hide comments

Works on: Instagram, Facebook, X, Threads

Moderation without the drama. A hidden comment disappears for everyone except the commenter: they still see it as posted, so there’s nothing to re-post and no “why did you delete my comment?” escalation. Spam, link-droppers, and bait go invisible without feeding them.

Hiding is fully reversible. Deleting is not: inbox:delete-comment removes the comment permanently, and the commenter can tell. Default to hide; keep delete for the genuinely bad stuff.

Reply to messages

Works on: Instagram, Facebook, X

DMs are where comment funnels convert and warm leads go quiet if you’re slow. Every conversation across your connected accounts is listed in one place; read a thread, reply, mark it read:

On Instagram and Facebook, DMs support link buttons and quick replies, and delivery and read receipts flow back into your stats. Like comments, message auto-replies run server-side from the Creator OS app: set the voice and links once and the inbox answers itself, with Kairos escalating anything sensitive instead of winging it.

Hide messages

Works on: Instagram, Facebook

Inbox zero for DMs: archiving a conversation hides it from the active inbox without deleting anything. The history stays intact and the thread comes back (or can be unarchived) if the person writes again. It’s how you keep the inbox to the conversations that still need a human.

Post to all platforms

Works on: Instagram, TikTok, YouTube, X, Facebook, LinkedIn, Threads

One flow for everything: upload media once, validate, then a single posts:create targeting every account that should get it. One call, seven platforms.

What you can publish where

Before big sends, validate: validate:post-length checks your caption against every platform’s character limit (Threads caps at 500, X at 280 free, the #1 cross-posting failure), and validate:media checks formats, sizes, and aspect ratios. Creator OS handles the per-platform payload differences for you, including TikTok’s required consent flags.

If one platform fails inside a multi-platform post, the others still publish. Check posts:get <postId> for the platform error, fix it, then posts:retry <postId>. No re-uploading, no re-posting the ones that worked.

Schedule everywhere

Works on: Instagram, TikTok, YouTube, X, Facebook, LinkedIn, Threads

Scheduling is the same posts:create call plus a time. Everything publishes from Creator OS servers at the scheduled moment; nothing keeps running on your machine. Schedule a month and close the laptop.

Whole-calendar work goes through posts:bulk-upload with a JSON file: one call schedules the entire calendar. Always --dryRun first; it validates every row (asset exists, caption fits, date is in the future) without posting anything.

The queue is inspectable and editable until the moment a post goes out: posts:list shows everything scheduled, posts:update edits a scheduled post, posts:delete pulls it. Ask analytics:best-time for the posting windows your accounts actually perform in, and schedule into those.

Post analytics

Works on: All connected platforms

Per-post performance across every platform in one place: impressions, reach, likes, comments, shares, saves, clicks, and views, per post, comparable across platforms. The same numbers the app’s Analytics tab shows are readable by your agent:

The point isn’t reporting, it’s judgment. An agent that reads last week’s numbers before scheduling next week doubles down on what worked: which hook, which format, which platform, which time slot. Ask “what performed best last month?” and the agent maps it to these commands.

Follower analytics

Works on: All connected platforms

Follower counts across every connected account in one call. The growth number that actually matters, tracked over time rather than eyeballed:

Instagram goes deeper: a daily follower-history series with per-day gained and lost (Creator OS snapshots this daily; Instagram itself no longer exposes a historical series), plus audience demographics by age, city, country, and gender once the account has 100+ followers.

Gained-vs-lost is the diagnostic pair: growth with high churn is a content problem, flat growth with no churn is a reach problem. The Creator OS dashboard draws this as the follower heatmap; your agent can read the same series and tell you which posts the spikes line up with.

Launch the Kairos dashboard

The Kairos repo ships with a local web dashboard for watching what your agent is actually doing, and verifying it’s working. From the kairos folder:

It opens at http://localhost:4180 (override with KAIROS_DASHBOARD_PORT). After npm link you can also launch it from anywhere with kai dashboard. Zero external services: it reads the repo’s files, the agent’s structured activity log, and the Creator OS API with your already-configured credentials. Missing credentials never crash it; you get a friendly connect state instead.

Every panel is fed by plain local JSON endpoints (/api/health, /api/activity, /api/automations, /api/chat…), and a panel is just one ES-module file in dashboard/public/panels/ with no build step, so you can add your own views against the same data.

CLI reference

Everything below runs as creatoros … (from the Kairos codebase, or via npx @creatoros/cli after init). Add --pretty to any read command for human-friendly output; agents usually consume the default JSON.

Setup, accounts & publishing

Inbox: comments & messages

Analytics

Scheduled agent runs

Anything not listed passes straight through to the underlying engine. Run creatoros --help for the full surface.