SOLVED
Your AI is part of the team
Agent jobs in Forge trigger on content events or a schedule. They run inside your existing setup, are managed like content, and stop when you say so. No separate service. No deployment.
Get started
No dashboard required.
Your AI handles publishing through a structured API. Your team stays in the tools they already use.
Full control, zero risk.
Role-based access means your AI can only do what you allow. Draft content never leaks.
Live in minutes.
Connect once. From that point, a conversation is all it takes to publish.
Most automation setups live outside the system they automate. A separate process, a hosted platform, a webhook receiver pointed at your API. You configure it elsewhere, deploy it separately, and hope the two stay in sync. When something breaks, the failure is usually at the seam between them.
Forge takes a different approach. Agent jobs run inside your Forge process. They trigger on content events or a schedule. They write results back through the same content API your team already uses. Nothing to deploy separately. Nothing to keep in sync.
Two ways an agent job triggers
When something is published. When a post goes live on forge-cms.dev, an agent reads the full post body and creates four draft social posts -- two for X, two for Mastodon, from two distinct angles -- for review. By the time you close the publishing tab, the drafts are waiting. You review them, queue what looks right, and move on.
On a schedule. A job runs at a set time, independent of any content event. Check a feed, process a result, send a summary. The agent works whether or not anyone is at a keyboard.
Managed like content
Agent jobs follow the same lifecycle as everything else in Forge.
A job in draft exists but does not run. Publish it and it becomes active -- it fires on its next trigger. Archive it and it stops, permanently and cleanly.
This means you control what runs, when it runs, and when it stops from the same place you manage the rest of your content. No config file change. No server restart. No deployment.
Going on leave? Archive the jobs that should not run without you. Back and ready? Publish them again.
The output is governed
Results the agent writes back -- draft posts, scheduled content, updated records -- go through the same content lifecycle as anything a human creates. Draft first. Published when approved. Archived when done.
The agent does not get a special path around the rules. It works within them, the same way your team does.
*See Signal bus for how content events trigger agent jobs.* *See forge-agent for wiring and configuration.*
Ready to put your AI to work?
Forge is open source and self-hosted. Get started in minutes.