Spektacular's pluggable architecture lets specs, plans, knowledge, and agents come from wherever your team already keeps them.
Two plugin points sit at the core of Spektacular. Implement either interface and register your backend to use it.
One interface, three use-cases: where specs live, where plans live,
and where knowledge is read from. Each section of config.yaml
independently names a Store provider, so you can mix and match.
The coding agent Spektacular drives during the implement workflow. Each registered agent declares its name and how to install workflow artefacts into a target project.
Plugins registered in the current release.
The file-system Store provider. Backs spec, plan, and knowledge
sections of config.yaml. Reads and writes plain markdown under a
configurable directory.
Drives Claude Code via
its skills and slash-command directory layout. Selected by
spektacular init claude.
Drives the Bob agent via its .bob/skills and .bob/commands layout.
Selected by spektacular init bob.
Drives the Codex agent via its skills directory layout. Selected by
spektacular init codex.
Concrete integrations the pluggable architecture is being built toward. Not yet shipping — listed here so teams can plan ahead.
Knowledge Store backed by an Obsidian vault. Lets teams point Spektacular at the markdown they already curate.
Knowledge Store backed by a Notion workspace. Surfaces team docs without copying them into the repo.
Spec Store backed by Jira. Tickets become spec inputs; planned work links back to the ticket they came from.
Implement the Store or Agent interface and register it. The Extending
page describes both.