Today we're launching TaskFlow Queue — a hosted async job queue built specifically for AI agents and long-running workflows.
Why we built it
Every team building with AI runs into the same set of problems. Model inference is slow and unpredictable. API rate limits interrupt your pipeline at the worst times. Failures are hard to debug because the work happens asynchronously across multiple systems.
General-purpose queues like SQS, Sidekiq, or BullMQ solve some of this — but they require significant infrastructure to set up and operate. We wanted something that worked out of the box for AI workloads: enqueue a job, provide a callback URL, get the result delivered when it's ready.
How it works
TaskFlow Queue is a hosted service. You send jobs via a simple HTTP API — a job is just a payload and a callback URL. We store it, process it, and POST the result to your callback when done.
We handle retries automatically (up to 3 times on Free, 10 times on Pro/Team), with configurable backoff. Failed jobs go to a dead-letter queue. Every callback is signed with HMAC-SHA256.
Who it's for
If you're calling AI APIs — OpenAI, Anthropic, Replicate, or others — and you need those calls to be reliable, retried on failure, prioritised by urgency, and delivered without managing worker infrastructure, TaskFlow is for you.
The Free tier gives you 1,000 jobs per day with no credit card required. Pro at $9/mo scales to 50K jobs/day. Team at $29/mo handles 500K jobs/day.
What's next
We're working on a dashboard with job history, dead-letter queue inspection, and webhook delivery logs in real time. Scheduled jobs and streaming callbacks are on the roadmap too.
Register for free and enqueue your first job in under five minutes.