When one model runs out (keeping a session going)
Some Claude models have their own usage limit, separate from your overall 5-hour and weekly meters. When one of them runs out mid-session, Omniscio can keep the session going on the backup model you already chose for rate limits, and return it to your own model once the limit resets. It ships off: until you turn it on, a single inbox card asks whether to switch one session or every session.
The short version
Your account's overall usage and a single model's usage are two different meters. You can be at 43% overall and still be out of Fable — and a session running Fable stops while Opus and Sonnet sit there unused.
Settings → Accounts → "Keep working when one model runs out" turns on the fix. It ships off.
What it does when it is on
- The session that ran out continues on the backup model you already set for rate limits, keeping its conversation — nothing restarts from scratch.
- It says so in the session, so you can see the model changed and why.
- It goes back to your own model once that limit resets.
- It does nothing when your overall usage is also out — changing models would not help there — and nothing twice in the same episode, so a session never bounces between models.
What it does when it is off
Nothing switches on its own. Instead you get one inbox card for that session:
- Switch this session — continues that one session on the backup model.
- Always do this — turns the setting on for every session from now on.
One card per session per model, so a session that keeps retrying cannot fill your inbox.
Questions you might have
Which model is the backup? The same one the rate-limit downgrade uses. If the backup is in the same family as the model that ran out, or is itself capped, Omniscio will not switch to it.
Does it cost more? It uses the backup model's normal pricing. Nothing here buys extra usage.
What if I picked a model for that session myself? The session still switches, and it still returns to the model you picked afterwards.