QStash alternative
An HTTP task queue you can also run yourself.
QStash got the model right: your route is the consumer, and the queue calls it, with retries, delays and schedules. Workly works the same way, but it's open source. Run it on your laptop, next to your services, or hosted at workly.run, and when a task fails, see every attempt and retry the whole group once it's fixed.
Same model. You choose where it runs.
QStash runs only on Upstash. Workly is one Apache-2.0 server, and the API, SDKs and dashboard are the same wherever it runs.
On your laptop
workly dev runs the production server on SQLite, with the dashboard. No tunnel, no license limits.
In your infrastructure
One binary plus Postgres, next to your services, delivering to private addresses too. Free, forever.
Hosted
workly.run runs the same server. Moving between them means changing WORKLY_URL and the API key.
Workly and QStash compared.
Both push HTTP requests, deliver at least once, sign every request and deduplicate on enqueue. The differences are in where it runs, where the settings live, and what's kept afterwards.
| QStash | Workly | |
|---|---|---|
| Delivery | HTTP push to your endpoint | HTTP push to your endpoint |
| Runs on | Upstash's service, in its EU or US region | Your laptop, any cloud or cluster, or hosted at workly.run |
| Source | Proprietary | Open source, Apache-2.0 |
| Local development | qstash dev: an in-memory server, licensed for development and testing only | workly dev: the same server as production, on SQLite, with the dashboard |
| Endpoints | Public URLs; locally, the dev server or a tunnel | Self-hosted, anything it can reach, private networks included. Hosted: public URLs |
| Settings | Per message, in headers: retries, delay, timeout, flow control | Per queue, in workly.yaml or the API; a task can override attempts and timeout |
| Retries | 4 attempts by default (3 retries), exponential up to a day, or your own formula; honours Retry-After | 5 attempts by default, backoff with jitter; honours Retry-After; you choose whether 4xx is retried |
| Rate limits | Flow Control: rate per period and parallelism, per key you choose | rate_per_second and max_concurrency per queue |
| Ordering | Queues deliver one at a time, in order | No ordering guarantee; set max_concurrency: 1 for one at a time |
| After the last retry | Dead letter queue, kept 3 days to 3 months by plan | Kept as failed with every attempt, grouped by error; an alert by email or Slack; retry one or a whole group once the fix is in |
| Callbacks | Per message: a callback with each response, and a failure callback | Per task or schedule: a callback with the outcome each time it finishes; the response stays in its attempt history |
| Deduplication | Deduplication ID or content hash, for 10 minutes | idempotency_key, for as long as the task is kept, and a stable Workly-Task-Id header |
| Schedules | Cron, UTC unless prefixed with CRON_TZ= | Cron with a timezone, overlap policy, run history and run-now |
| Fan-out and fan-in | URL Groups send one message to many endpoints; fan-in is in Upstash Workflow | Task groups: enqueue many, get one callback with the counts when all have finished; workflows can wait for one before the next step |
| Workflows | Upstash Workflow, a separate product: steps written as code in one function, with sleeps, waits for events and results between steps | DAGs of HTTP steps in workly.yaml or the API, each step a task in your queues; retry a failed run from where it stopped |
| Visibility | Console logs with responses, kept 3 to 14 days by plan; Prometheus and Datadog with Prod Pack (+$200 a month) | Dashboard with every attempt, failures grouped by error, MCP that can retry and pause, and Prometheus metrics when self-hosted |
| Request length | 15 minutes on the free plan, 2 hours pay-as-you-go, up to 12 hours on fixed plans | Up to 30 minutes per attempt |
| Price | 1,000 messages a day free, then $1 per 100k; each retry counts as a message | Free to self-host. Hosted: 100k tasks a month free; a task counts once, however many attempts |
QStash behaviour and limits from Upstash's documentation and pricing page, November 2026.
Settings move from each message to the queue.
With QStash, every publish says how to retry and how fast to deliver. With Workly, the queue says it once, in a file you deploy with your app, and every enqueue to it follows. Change the rate in one place, not in every caller.
import { Client } from "@upstash/qstash";
const qstash = new Client(); // QSTASH_TOKEN
// Every call carries its own settings
await qstash.publishJSON({
url: "https://api.example.com/tasks/send-email",
body: { userId },
retries: 5,
deduplicationId: `welcome-${userId}`, // for 10 minutes
flowControl: { key: "emails", rate: 5, parallelism: 10 },
});import { Workly } from "@worklydev/sdk";
const workly = new Workly(); // WORKLY_URL, WORKLY_API_KEY
// The queue holds the settings; the call names it
await workly.tasks.enqueue("emails", {
url: "https://api.example.com/tasks/send-email",
payload: { userId },
idempotencyKey: `welcome-${userId}`, // for as long as the task is kept
});# workly.yaml, applied with `workly deploy`
version: 1
queues:
emails:
max_attempts: 5
rate_per_second: 5
max_concurrency: 10And schedules know what ran.
Both run cron in your timezone. Workly also decides what happens when last night's run is still going, keeps each run with its attempts, and has a button to run one now.
await qstash.schedules.create({
destination: "https://api.example.com/tasks/nightly-report",
cron: "CRON_TZ=Europe/Stockholm 0 3 * * *",
});# workly.yaml
version: 1
schedules:
nightly-report:
cron: "0 3 * * *"
timezone: Europe/Stockholm
url: https://api.example.com/tasks/nightly-report
overlap: skip # if last night's run is still goingSee why a task failed, then fix them all at once.
Every task keeps its attempts: status code, error, response body and timing. Failed tasks are grouped by error, so after a bad deploy you fix the handler and retry the whole group, from the dashboard, the CLI or your coding agent over MCP. Queue depth, delivery rate and the age of the oldest due task are in the dashboard too.

When QStash is the better choice.
You need an SLA and a SOC 2 report today
With Prod Pack, QStash has a 99.99% uptime SLA, a SOC 2 Type 2 report and encryption at rest, in an EU or a US region. If a security review needs those now, QStash has them.
You need what Workly does not have
Ordered queues, URL Groups that fan one message out to many endpoints, the response body in each callback, and requests that run for hours. Workly has none of these; its callbacks report the outcome, and its attempts stop at 30 minutes.
Volume is small or spiky
Pay-as-you-go pricing scales to zero: a quiet month costs cents. Hosted Workly is free up to 100k tasks a month, but beyond that it is a monthly plan.
Questions
- Both deliver HTTP requests. What is the difference?
- Where it runs and what you see. QStash runs only on Upstash. Workly is open source: run it next to your services, on your laptop with the same server as production, or hosted at workly.run. And Workly keeps every attempt of every task, groups failures by error and lets you retry a group once the fix is in.
- Can I self-host QStash?
- No. The QStash CLI runs a development server, but its license allows only development, testing and CI, and it keeps everything in memory. Workly is the same Apache-2.0 server everywhere: in development, self-hosted in production, or hosted.
- Does Workly work from Vercel, Cloudflare Workers or Lambda?
- Yes. Enqueueing is one HTTP request, and the TypeScript SDK runs anywhere with fetch. Your handlers are ordinary routes, and the signature check works with any Fetch API request. Keep each attempt under the platform’s own time limit, and Workly’s 30 minutes.
- What about Upstash Workflow, Inngest or Trigger.dev?
- Those run multi-step functions written as code, with sleeps, waits for events and state passed between steps. Workly's workflows are a DAG of your endpoints declared up front: each step is an ordinary task, a step can fan out over a task group, and a failed run retries from where it stopped, but there are no conditions or outputs between steps. If your steps branch or carry state, those tools fit better.
- How much changes in my handlers?
- Little. Both deliver a POST to your route and expect a 2xx. Swap Upstash's signature check for Workly's verifySignature, and use the Workly-Task-Id header where you used Upstash-Message-Id to make handlers idempotent.
- How do I move over gradually?
- Endpoint by endpoint. Create a queue for one, enqueue to it with Workly instead of publishing to QStash, and accept both signatures on the route while messages already in QStash drain. Move schedules once their tasks are on Workly.
Try it in five minutes.
pip install "workly[cli]" and workly dev give you a server and dashboard on your laptop, with nothing else to start. Or sign in to hosted Workly and get a workspace and an API key in a minute, free on the Hobby plan.