Skip to content

Quickstart

Everything here runs on your machine - no server, no account.

Terminal window
curl -fsSL https://hyperspool.com/install.sh | sh

One static binary, spool: the CLI, a local dev server, and the agent.

Terminal window
spool init && spool run pipeline.yaml

spool init writes three starter workflows; spool run spins up an ephemeral in-process instance, runs the pipeline, streams the output, and tears everything down. The exit code mirrors the run.

pipeline.yaml is a CI-shaped pipeline with a durable timer between build and ship. Make that timer 30d and it still works - the run parks without holding a machine and wakes on time. (For a wait you intend to outlive the terminal, use spool dev below - a one-shot spool run instance is ephemeral and takes its parked runs with it.)

Terminal window
spool run durable.sh

durable.sh is plain bash. It draws a random number in a spool step, sleeps, then adds one to it - and prints the same number on both sides of the sleep.

Watch the output: the script runs twice. spool sleep ends the process rather than pausing it, so the wake-up starts from the first line again. The number doesn’t change across those two passes, because a spool step records its stdout the first time and replays the recording afterwards. That is the whole trick: values survive the gap, so the code after a sleep can use what the code before it produced. The gap can be thirty days, and the second pass can land on a different agent on a different machine.

Terminal window
spool run signals.sh

This one parks on spool wait-signal. When it does, it prints the exact command to send the signal from a second terminal; the payload becomes the command’s stdout and the run completes.

Terminal window
spool dev

Starts a single-process Hyperspool - server, agent, sqlite, and the web UI - on http://127.0.0.1:18753. Other spool commands in the same account find it automatically, secrets persist across restarts, and the dashboard shows runs waiting on signals or approvals.