Provider retry
Retry turns after transient provider errors and when subscription limits reset.
Overview
View sourceKeep threads moving when a provider is overloaded or unreachable, or your subscription window is used up. This plugin watches for failed turns and schedules a retry at the right time. You do not have to come back and press send.
What you get
- Automatic retry after a provider overload, with a short delay that doubles on each attempt.
- Automatic retry after a server error or unreachable provider (HTTP 500, 502, 503, or 504, or a dropped connection), starting at one minute and doubling on each attempt.
- Automatic retry after a subscription limit, timed to the reset the provider reports.
- A queued message card on the thread that you can send now or cancel.
- A cap of four retries per turn.
- No retry of a scheduled run of a recurring automation. That automation runs again on its own schedule.
How it works
The plugin reacts to each failed turn. It retries overload errors, server and connection errors, subscription-window rate limits with a known reset time, and rate-limit failures routed through Account Pooler when the pool reports a usable account or a known recovery time. Credit and spend limits are not retried. Each retry waits a little past the reset and adds a random spread. Many threads on one account then do not wake at the same instant. Messages that fail while a retry is already waiting join it instead of adding another retry turn, unless the provider never received them.
Settings
Maximum automatic wait: skip a retry when the reset is farther away than 6 hours, 24 hours, or no limit.
CLI
bb provider-retry status [thread-id] [--json]: show pending retries.bb provider-retry retry <thread-id>: send a pending retry now.bb provider-retry cancel <thread-id>: cancel a pending retry.
For pooled threads, the pool’s availability overrides provider-local quota snapshots, which can be missing or refer to an account the pool has already replaced. Recovery uses the earliest usable account after all its blocking windows and temporary holds clear, including limits for the failed request’s model. Parent pools use the same authenticated availability lookup. Unknown resets, authentication failures, and unavailable sources are explained without scheduling a speculative retry.
Run bb provider-retry explain [thread-id] [--json] to inspect the last automatic decision, including skipped retries. This is a historical decision, not the current queue state; use status for pending retries. Decisions are recorded after this version is loaded and survive restart. The same diagnostic is available through bb.sdk.plugins.callRpc using plugin provider-retry, method decision.get, and input { threadId }.
More from BB
More in Agents & Providers 85
Add, choose, configure, route, or coordinate who runs a thread.

