Runspore documentation
Use the spore CLI to create workflows, run and inspect work, handle signals, and recover interrupted runs. Start here, then use the SWL standard when you need the language reference.
Use Runspore through the spore CLI: write a workflow, start a run, inspect what it needs, and keep it moving until it finishes. These pages describe the implemented CLI and workflow dialect. You do not need to read the SWL standard or the runtime's architecture before using it.
Start here
- Getting started: install
sporeand run a complete example from an empty folder. - Authoring workflows: commands, input and output, decisions, signals, retries, loops, validation, and artifacts.
- Running workflows: choose a driver, inspect progress, send answers, recover interrupted work, and cancel.
Common tasks
| Task | Guide |
|---|---|
| Let a person, script, or agent perform a step | External and asynchronous steps |
| Drive Runspore from scripts or CI | Automation and JSON output |
| Find a run, understand its database, or configure a worker | Projects and settings |
| Look up a command, flag, or exit code | CLI reference |
| See progress across runs and get a notification | Watch a run |
| Use Claude Code in a workflow | A Claude step |
| Fix a workflow that does not validate | Validation guide and diagnostic catalogue |
| Try an existing workflow | Example catalogue |
SWL standard
The SWL chapter covers the independent Spore Workflow Language working draft. Its public source defines portable contracts for implementations. It is a reference chapter, not an installation or Runspore command guide. Runspore's current workflow.dot dialect is implementation-specific; direct execution of SWL notation is not yet claimed.
Which documentation applies?
These guides use the current runspore/0.2 authoring dialect and runspore.cli/0.2 command contract. Check spore --version and spore capabilities for the software you installed; a software release such as 0.9.1 does not imply a workflow format named runspore/0.9.1.
For exact field definitions and behavior, use the implementation specifications. The architecture and design documents also describe future functionality. Neither those plans nor a feature in the SWL draft means the current CLI implements it.
Runspore currently supports one machine, one OS user, and SQLite. Workflows can execute programs with that user's access. Keep workflow source under version control and keep run data out of it; Projects and settings explains where both live.