Workspace Templates

Workspace templates let you export your full SwarmClaw setup — agents, skills, schedules, connectors, chatrooms, MCP servers, projects, and goals — into a single JSON bundle that you can import into a fresh install. Useful for:

onboarding a new install with your existing setup in one click
backing up your configuration before risky changes
sharing a working setup with someone else (credentials are scrubbed)

What's Exported

The bundle is portable JSON, schema version 2. v1 bundles produced by older versions still import.

Category	Notes
Agents	All non-trashed, non-disabled agents. Credential IDs, gateway IDs, runtime spend fields, and thread-session IDs are stripped.
Skills	All skills with full content.
Schedules	All non-archived schedules. Re-imported as `paused` so you can review before they fire.
Connectors	All connectors. Tokens/keys/passwords/credentials are scrubbed (any field whose key matches `/(token
Chatrooms	Non-archived, non-temporary rooms with their routing rules and member agent IDs (remapped on import).
MCP Servers	Transport, command, args, URL preserved. Env and header keys preserved with empty values; you fill them after import.
Projects	All projects with their objectives, audiences, priorities, and metrics.
Goals	Full goal hierarchy. Parent-goal references and project/agent references are remapped to the new IDs on import.
Extensions	Manifest reference only (name, source, version). Extension code is not bundled — re-install from the extension marketplace or source URL.

What's Not Exported

Credential values, OAuth tokens, API keys (anything that looks like a secret)
Trashed agents
Sessions, messages, usage records, mission runs, activity log
Generated avatars, thread sessions, runtime caches

Export

GET /api/portability/export

Returns the full manifest as JSON.

Import

POST /api/portability/import
Content-Type: application/json

{ "manifest": { ... } }

Returns a summary listing what was created, skipped, and which records need credentials added before they can be enabled:

{
  "agents":     { "created": 8, "skipped": 0, "names": [...] },
  "skills":     { "created": 12, "skipped": 0, "names": [...] },
  "schedules":  { "created": 3, "skipped": 0, "names": [...] },
  "connectors": { "created": 4, "skipped": 0, "names": [...], "needsCredentials": ["Discord (Main)", "Slack (Ops)", ...] },
  "chatrooms":  { "created": 2, "skipped": 0, "names": [...] },
  "mcpServers": { "created": 5, "skipped": 0, "names": [...], "needsCredentials": ["GitHub MCP"] },
  "projects":   { "created": 3, "skipped": 0, "names": [...] },
  "goals":      { "created": 7, "skipped": 0, "titles": [...] },
  "idMap":      { "<originalId>": "<newId>", ... }
}

Names that conflict with existing records get a (2), (3), … suffix automatically.

CLI

swarmclaw portability export > my-workspace.json
swarmclaw portability import --json @my-workspace.json

Limits

Extensions are referenced by manifest, not bundled. Re-install them on the target machine.
Connectors and MCP servers re-import disabled with credentials cleared. Add credentials in the UI, then enable.
The default workspace assumption: imports add to the active workspace. Multi-workspace data isolation is on the roadmap.