Live demo deployment

Run it now.
Run it again later.

loops.yaml is a small scheduler for shell commands and recurring Pi or OpenCode prompts. Define one loop, run it on demand, or leave it watching on a schedule.

Try it now Get the source

RuntimeBun + Pi / OpenCode LicenseMIT Hosted onCloudflare Workers

About this deployment

This is the project’s public demo deployment and documentation. The scheduler itself runs wherever you install it; the complete implementation is in the source repository.

Browse the repository ↗

01 / What it is

One loop.
Two execution modes.

The same idea works for unattended commands and active agent sessions: say what should happen, say when, and keep the execution visible.

Shell scheduler

Commands that run again.

Use a short YAML file for backups, health checks, deploys, and any command with an exit code.

loops watch

Agent extension

Recurring prompts while your agent is open.

Set an interval and a stop condition in Pi or OpenCode. The agent waits for idle time before it starts the next check.

/loop every 30s …

Run record

Logs and status stay visible.

Every shell run writes a timestamped log and latest status. Pi loops remain attached to their session.

loops logs review

02 / Define

Define the schedule in one readable file.

loops:
  review:
    schedule: "0 8 * * *"   # omit for on-demand only
    run: ./scripts/review.sh

03 / Run

loops listsee every loop and its latest run
loops run reviewrun one command now
loops watchkeep scheduled commands running
loops logs reviewread the latest output

04 / Try it now

Install it and go.

Install the scheduler from the public repository — no hosted account required. Then run a loop on demand, or leave it watching.

Shell scheduler

bun add -g github:acoyfellow/loops-yaml

Shell quickstart ↗

Prefer a coding agent?

Run the same loops as recurring prompts inside Pi and OpenCode today — with more agents on the way.

See all runtimes →

Small surfaceFour CLI commands.

Enough interface to run, watch, list, and inspect.

Plain stateState stays in files.

YAML definitions, timestamped logs, and JSON run state.

Visible boundaryProcess-level access.

Each loop uses the permissions of the process that starts it.

Source available

The complete project
is in the repository.

Review the implementation, tests, reference documentation, Pi and OpenCode extensions, and website source.

Open the source Browse working recipes →

Run it now.Run it again later.

One loop.Two execution modes.