Account, billing & API

Jobs, errors, and status

Every generation in Framesail runs as a background job: kick it off, leave the page, come back. Each project reports its jobs' type, status, and any errors, and a step-by-step progress report shows where the pipeline stands — so "is it still running?" always has an answer in the UI (or via get_workflow_status and await_jobs over MCP).

When a job fails

• ·Errors surface to the UI on the job itself — you'll see what failed and why, not a silent stall.
• ·Provider quota-style errors fail fast rather than hanging.
• ·A watchdog flips genuinely stuck jobs to an error state, so nothing blocks the pipeline forever.
• ·The fix is almost always to retry the failed item — and retries are safe by design, below.

Retries never double-bill

Duplicate submissions are handled server-side: single-run steps refuse to start while already running, and batch steps (references, voice blocks, segment renders) skip items already generated or in flight. Insufficient credits are caught up front — a job you can't afford fails at submission, before anything runs or is charged (see plans & credits).

Step-order errors

Calling a step before its prerequisites are done fails with a step_not_ready error that names the exact next action — the pipeline's unlock ladder is enforced server-side for the web app, API, and MCP alike. The ladder itself is described in make your first video.

Rate limits

Generation endpoints are rate-limited. Normal interactive use won't hit the limits; scripted bursts against the API might, and should back off and retry.

For crawlers and agents

The site itself is agent-readable: /llms.txt lists the pages that matter, content is served as markdown via content negotiation, and Link headers advertise the API catalog.