Vault Inc
Start free

Capabilities

QueuesPush work out of the request and answer straight away SchedulesCron expressions with a timezone that behaves Retries and backoffPer-job policies, and a dead letter queue you can drain Run historyEvery attempt, its arguments and its output

Built for

Email and webhooksDeliver on someone else's schedule, not yours Nightly pipelinesLong imports that must not overlap Customer exportsWork you have to be able to prove ran

New in 2.7

Concurrency keys cap how many runs of a job execute at once, per customer.

Read the changelog

Learn

How Vault Inc worksThe four moving parts, in order Retry patternsPicking a backoff that suits the failure Common questionsLimits, billing and what counts as a run

Keep up

ChangelogEverything shipped, most recent first PlansWhat each tier includes

Changelog

What shipped, and what it broke.

Every release, newest first, including the fixes we would rather not have needed. Breaking changes get a version of their own and six months of notice.

2.7 30 August 2026 Feature

Concurrency keys

Cap how many runs of a job execute at once, keyed on anything in the arguments. The obvious use is one import per customer at a time; the less obvious one is protecting a third-party API that rate-limits per account rather than per token.

  • concurrency.key accepts any JSON path into the job arguments
  • Runs beyond the cap wait in order rather than failing
  • The run list shows what a job is waiting behind
2.6 12 August 2026 Improvement

Faster pickup, quieter workers

Median time from enqueue to pickup is now 38ms, down from 210ms, because workers hold a long poll instead of asking every two seconds. Existing workers get this by upgrading the client; nothing changes in your job code.

  • Long polling replaces the fixed 2s interval
  • Roughly 90% fewer requests from an idle worker
  • Client libraries 2.6.0 and above
2.5 28 July 2026 Fix

Schedules and the clocks going back

A schedule set to 01:30 Europe/London fired twice on the night the clocks went back and not at all when they went forward. Both are fixed: an ambiguous local time now fires once, on the first occurrence, and a skipped one fires at the next valid minute.

  • Affected schedules with a non-UTC timezone and a fire time inside the transition hour
  • No action needed; the fix applies to existing schedules
2.4 9 July 2026 Feature

Drain the dead letter queue

Dead letters can be sent back to the main queue in bulk, filtered by job, date range or error, with the original arguments intact. Previously the only route back was one at a time through the API.

  • Filter, preview the count, then drain
  • Drained runs keep a link to the attempt that failed
  • Available on every plan
2.3 24 June 2026 Improvement

Run history you can search

Index any argument as a searchable field and filter the run list on it. Most people index a customer or tenant id, which turns "did this run for them" from a support ticket into a URL.

  • Up to five indexed fields per job
  • Applies to runs created after the field is added

Nothing here yet that you need? Tell us what is missing.

Roadmap requests go straight to the people building it.

Start free Read the docs
Back to GSAP Vault