Skip to main content
MCP resources are cacheable reads — pure GETs against a pleyor:// URI. They’re how the assistant inspects state without spending a tool call.

Discovery resources

URIReturns
pleyor://nodesLists every node type Pleyor supports (~47), grouped by category, with ports + descriptions.
pleyor://nodes/{type}Full definition for a single node type — config schema, port shapes, capability kind, validation rules.
pleyor://modelsThe model catalog — every model variant Pleyor routes to, with its capabilities and pricing.
pleyor://models/{variant_id}One model variant with full settings.
Read pleyor://nodes once at the start of any build session — the assistant needs it to pick node types.

State resources

URIReturns
pleyor://agentsList of agents in the bound organization.
pleyor://agents/{id}One agent — current graph + latest version metadata.
pleyor://agents/{id}/versionsVersion history for an agent.
pleyor://appsList of published apps.
pleyor://apps/{id}One app’s metadata + input schema.
pleyor://runs/{id}A run’s full snapshot — top-level status, per-step status, per-step outputs. Used by wait_for_run internally; you can poll it manually if you want progress mid-run.

Account resources

URIReturns
pleyor://assetsRecent assets the organization has produced.
pleyor://brand-kitsBrand kits available in the organization.
pleyor://brand-kits/{id}One brand kit’s full content.
pleyor://integrationsConnected external accounts (Google Drive, scrapers, etc.).
pleyor://creditsCurrent credit balance + recent transactions.

Reference resources

URIReturns
pleyor://skills/workflowsThe pleyor-workflows agent skill — the canonical playbook for build/run/iterate. Server instructions direct new agents here. Always reflects the latest skill version because it’s served straight off the MCP server (no separate CDN to drift).

UI resources

URIReturns
ui://pleyor/workflow-canvasHTML shim that the chat client’s MCP-Apps sandbox loads to render the in-chat canvas.

When to read what

The workflow skill opens with the “Discovery — always do this before designing” section that maps tasks to the resources you should read first. Short version:
  • Building a new workflow → read pleyor://nodes, then pleyor://nodes/{type} for each node type you’ll use
  • Editing an existing agent → read pleyor://agents/{id} (or call view_agent)
  • Picking a model → read pleyor://models or call suggest_model
  • Checking a run → call wait_for_run (single round-trip) instead of polling pleyor://runs/{id}
  • Loading a brand context → read pleyor://brand-kits then pleyor://brand-kits/{id}

See also

Tools

The 20+ tools that mutate or trigger state.

Errors

What RESOURCE_NOT_FOUND and friends mean.