AI agents fail in predictable ways: wrong OAuth scope, exhausted monthly quota, channel not connected. Storylayer v1.2.0 addresses both discovery and recovery.
MCP resources
Hosted MCP at https://app.storylayer.ai/api/mcp now advertises resources agents can read without burning tool calls:
| URI | Content |
|---|---|
| storylayer://docs/agent-quickstart | Markdown onboarding |
| storylayer://docs/ai-catalog | Full ARD manifest JSON |
| storylayer://docs/openapi | OpenAPI 3.1 spec |
| storylayer://docs/gpt-instructions | Assistant-optimized instructions |
Resources appear in the MCP server card (/.well-known/mcp.json) with capabilities.resources: true.
Scope error hints
When a token lacks permission, MCP tools now return hints like:
Token missing scope(s): stories:publish. Fix: Add stories:publish scope to schedule or publish. Publishing requires connected channels — call list_social_connections first.
This cuts down on "why did publish_story fail?" loops in Claude and ChatGPT.
whoami quota fields
whoami now returns:
plan— free, pro, agencymcp_usage—used_this_month,remaining,limitdiscovery— URLs for ai-catalog, MCP card, quickstart, OpenAPI
Agents can self-check quota before a bulk create_stories_bulk run.
Try it
Connect any MCP client to https://app.storylayer.ai/api/mcp (OAuth or PAT). Call whoami first, then list resources in your client UI.
Docs: storylayer.ai/docs/mcp