AI agents
An AI agent is a helper you set up once. It writes draft posts for you again and again, on a schedule you pick.
Why use it
Instead of sitting down to write every post, you tell the agent what to post about and how often. It quietly prepares drafts for you. You stay in control: nothing goes out until you read it and approve it.
Step by step
Create your agent
Open the Agents page and click New agent. Give it a name and tell it what kind of posts you want, for example "friendly tips about our coffee shop."
Pick channels and how often
Choose which social accounts the agent should write for. Then set how often it should run, for example once a day in the morning, and how many posts to prepare each time.
Turn it on and let it run
Switch the agent on and save. From now on it will wake up on your schedule and write fresh drafts by itself. You can close the page and walk away.
Review what it wrote
When you come back, open the agent's drafts. Read each one. Edit anything you like, then approve and schedule the good ones. Skip the rest.
WARNING
An agent never posts on its own. It only prepares drafts. A post goes live only after you read it and approve it.
Good to know
- Every time the agent runs, it uses credits, just like when you write a draft yourself by hand.
- Each text draft costs 1 credit. An image costs 5 credits and a video costs 20 credits.
- You can set a daily limit so the agent never prepares more than you want in one day.
- You can turn an agent off any time, and it stops right away.
Related
For developers
Agents are stored as documents with a schedule_cron field (5-field cron, e.g. "0 9 * * *"), a timezone, target accounts, posts_per_run (1–10), and a daily_limit (1–50).
The ai-agents BullMQ queue runs a repeating job per agent (aiAgents.runner.ts). Each run loads the agent plus the user's last 5 posts for context, checks the daily limit, generates up to min(posts_per_run, daily_remaining) drafts through the Qevron gateway, and saves each as a Post with status: 0 (draft) and approval_status: "pending_review" — so nothing publishes without human sign-off.
Endpoints: GET/POST /api/v1/ai-agents, GET/PATCH/DELETE /api/v1/ai-agents/:id, POST /api/v1/ai-agents/:id/run (one-shot manual trigger that respects daily_limit), and GET /api/v1/ai-agents/:id/runs (run history).