Cloudflare adds per-Workflow concurrency caps, with limits developers need to account for
Cloudflare has added a way to limit how many instances of an individual Workflow run simultaneously. The November 10 change gives developers a more targeted control over workloads that share an account, including applications calling databases or external AI services. Instances beyond the configured cap queue rather than fail. Cloudflare announcement
The setting requires Wrangler 4.127.0 or newer. Developers must preserve it in every deployment configuration: deploying without it removes the cap. Reducing a limit also leaves already-running instances alone, so it is not an emergency stop for existing work. Release details
The documentation adds several operational qualifications. A cap does not reserve capacity, and the account-wide limit still applies. Short sleeps or retries can continue occupying slots. Local development does not enforce the limits. On Workers Paid, scheduled instances can bypass the cap during their cron concurrency budget, for up to an hour per firing. Concurrency documentation
For teams building AI workflows, the practical value is controlling bursts against shared dependencies. That is an infrastructure control, not a model upgrade or a guarantee of lower overall spending. Teams should test the production scheduling behavior rather than assuming a local run demonstrates how the cap works.
Sources
- Cloudflare per-Workflow concurrency announcement (primary; 2026-11-10)
- Concurrency limits documentation (primary; 2026-11-10)
Sources and citations
Revision history
Published .