Skip to main content

​
Integration Ecosystem

SI supports a two-layer integration model:
  1. Core integrations in the SI codebase.
  2. External plugin-pack catalogs maintained independently and loaded at runtime.

​
Why This Model

  • Faster integration iteration without waiting for SI core releases.
  • Namespace ownership (openclaw/*, saas/*, etc.) to avoid ID collisions.
  • Stable operator workflow using catalog artifacts (catalog/*.json) and SI_PLUGIN_CATALOG_PATHS.

​
External Plugin-Pack Repository

A sibling repository (../si-integrations) now contains:
  • OpenClaw parity integrations (all current OpenClaw extension plugin IDs).
  • SaaS-first integrations for startup/company workflows.
  • Deterministic catalog generation scripts and validation scripts.

​
Operational Workflow

  1. Add or update plugin manifests in the external pack.
  2. Build catalogs (npm run build:catalogs).
  3. Validate catalogs (npm test).
  4. Load into SI:
SI_PLUGIN_CATALOG_PATHS="/absolute/path/to/si-integrations/catalog/all.json" si plugins list --json

​
SI Catalog Tooling

SI now provides native catalog build/validate workflows for manifest trees: 
si plugins catalog build --source ./plugins --output ./catalog/all.json --channel ecosystem --tag integrations
si plugins catalog validate --source ./plugins --json
These commands make large-scale integration packs maintainable and testable with a single interface.
  • Keep each plugin namespaced and owner-scoped.
  • Use install.type=none for metadata-only catalog entries.
  • Attach channel/tags metadata in generated catalogs for discoverability.
  • Add policy controls in SI runtime (si plugins policy) before enabling at scale.