Clarifying Gamma Claude Connector API Usage, URL Patterns, Image Options, Layout, and Rate Limits

Hi — we use the Gamma Claude Connector programmatically (MCP tools generate, read_gamma, get_themes, get_folders). We've resolved most of our questions from developers.gamma.app, but 11 points remain ambiguous or undocumented. Where possible, please link the canonical reference alongside your answer. Generation URLs and status Does gamma.app/generations/{generationId} exist as a user-facing URL? If so, does it persist and/or redirect to gamma.app/docs/{gammaId} once rendering completes? We can't find this URL pattern documented anywhere, but your API responses use it. Does the Claude Connector's generate tool block until rendering completes, or return immediately while work continues async? Tool description phrasing ("share the gammaUrl when it's available") is ambiguous.

imageOptions.source enum drift 5. Four of your documentation surfaces publish different enum lists. Which is canonical? - REST API spec + parameters guide + llms-full.txt (10 values): ...pexels, noImages, themeAccent - MCP tools reference page (9 values): ...unsplash, placeholder, noImages (no themeAccent, no pexels) - Live Claude Connector schema (9 values): ...pexels, placeholder, noImages (no themeAccent, no unsplash) 6. How do we pass themeAccent via the Claude Connector? The connector schema rejects it. 7. Status of unsplash — has it been renamed to pexels? Deprecated? Still valid somewhere? Not in the changelog. Image layout 8. When a slide has multiple real image URLs of the same aspect ratio (e.g. three 9:16 vertical ad screenshots), does Gamma preserve the aspect ratio or crop to a uniform shape? 9. Can additionalInstructions reliably hint gallery layout (row of portraits, square grid, etc.) or is output always variable? Your charts/structured-content guide says layout output "varies between runs." Limits 10. What are the actual numeric rate limits (burst / hourly / daily) per plan tier? The headers are documented but the values aren't. 11. Can a generation silently stall in pending indefinitely on large payloads, or does it always resolve to completed or failed? The docs list three terminal statuses but don't mention timeout behaviour.