High-level ideas and open questions. Concrete work lives in plans/.
- Pull collections and globals to JSON files with virtual field stripping
- Push with conflict detection (manifest-based)
- Status (offline) and diff (online) commands
- CRUD commands: find, create, update, delete, count, versions, restore, duplicate, upload, request
- Lexical editing commands: list, get, add, replace, remove, set, link, search, diff
- Per-locale file model (
{id}_en.json,{id}_de.json) - Draft support on pull and push
- Block-aware
--fieldpaths for lexical commands - Payload plugin for schema endpoint
- Stdin support for lexical piping (
--json -) --wherefiltering on pull and CRUD commandsmecommand for auth verificationskillcommand for agent skill installation- Grouped
--helpoutput - Profiles for switching between CMS instances (
--profile,PAYLOAD_PROFILE)
- Human approval before push — commit/push model, push queue, or CMS-side review?
- Lexical to Markdown conversion
- Watch mode
- Versions/drafts in sync workflow — pull version history to files, diff between versions
- Multi-tenant support — base filters for tenant-scoped operations within a single instance (distinct from profiles, which switch between instances)
- Batch upload — bulk upload of media files (images, PDFs, etc.) in a single command (e.g. from a directory), with progress and error reporting
- Expose local content validation to agents — JSON Schema is consumed by IDEs today; agents can use
mcp__ide__getDiagnosticswhen an IDE is attached, but a push-side ajv check is needed for headless runs (seeplans/007-content-validation-for-agents.md)