Skip to main content
Scheduled agents are Fimo’s way of keeping a site fresh without a human in the loop. An agent is a small bundle of committable files in your repo — you deploy it like the rest of your code, then wire it to a trigger so it runs on its own. Because the definition lives in git, an agent is reviewable, diffable, and portable. The cloud only holds the activations (which triggers are wired on which environment) and the run history.

​
What an agent is

Every agent lives under agents/<name>/ in your project:
  • config.yaml — triggers, requested secrets, and an integration allowlist.
  • GOAL.md — what the agent should accomplish, in plain language.
  • Optional files such as capabilities.yaml for composed sub-agents.
Omitted integrations and tools are simply unavailable to the agent, so the allowlist is your safety boundary.

​
Why they matter

Autonomous upkeep

Translate new content, sweep for broken links, or run a post-deploy check — on a cron, unattended.

Versioned like code

The bundle reaches the server through fimo deploy, so agents branch, review, and roll back with everything else.

Scoped by design

An explicit integration and tool allowlist means an agent can only touch what you grant it.

Inspectable runs

Every run records a report and history you can read back from the CLI.
The bundle lives in git, not the database. Editing an agent locally is not enough — run fimo deploy to push the bundle before you activate or run it.

​
What’s next

Create an agent

Scaffold a bundle from a template, write its goal, and validate it.

Schedules

Wire an agent to a cron trigger and trigger runs on demand.

Runs, reports & logs

Inspect what an agent did on each run.