Create a queue, send a message, consume it from any worker, and debug retries, replay, and DLQ in one dashboard. One API key. Zero queue ceremony.
// ~60 seconds from npm i to first message + dashboard.
# install
$ npm i @asyncbase/sdk
// 3 lines: enqueue a message
import { Queue } from "@asyncbase/sdk"
const q = new Queue(process.env.ASYNCBASE_KEY!)
await q.send("emails", { to: "jane@example.com" }, { delay: "30s" })
// consume with auto-retry + DLQ — no Redis, no SQS, no IAM
for await (const msg of q.consume("emails", { group: "workers" })) {
try { await send(msg.payload); await msg.ack() }
catch { await msg.nack() }
}Not “HTTP push + write your own queue glue.” AsyncBase drops in as the queue layer for the patterns you already know — decorators, Laravel queue workers, Python async, Go context.
@Processor("emails")
export class EmailConsumer {
@Process()
handle(msg: Job) {
return sendEmail(msg.payload)
}
}# .env
QUEUE_CONNECTION=asyncbase
# your code, unchanged
SendEmailJob::dispatch($payload);import { Queue } from "@asyncbase/sdk"
const q = new Queue(process.env.ASYNCBASE_KEY!)
await q.send("emails", { to })from asyncbase import Queue
q = Queue(os.environ["ASYNCBASE_KEY"])
await q.send("emails", {"to": to})Also available: @asyncbase/sdk (Go), @asyncbase/cli.
None of the incumbents (QStash, SQS, Cloudflare Queues, Hatchet, Vercel Queues) have an MCP server. Your AI agents can send, inspect, and redrive queues without you writing a single adapter.
7 tools exposed: send, pull, ack, nack, list_dlq, redrive_dlq, get_docs. Point Claude Desktop / Cursor / Windsurf at npx @asyncbase/mcp and your agents talk to queues directly.
/asyncbase-setup auto-detects your stack (Next.js, Laravel, NestJS, FastAPI) and scaffolds env vars, SDK install, and a working consume loop. Zero copy-paste, zero hallucinated imports.
# Claude Desktop / Cursor / Windsurf config
{
"mcpServers": {
"asyncbase": {
"command": "npx",
"args": ["-y", "@asyncbase/mcp"],
"env": { "ASYNCBASE_KEY": "sk_live_..." }
}
}
}Researched, cited. This is what you actually hit.
| Option | What breaks for a side-project builder | Source |
|---|---|---|
| Apache Kafka | "Running Kafka in production demands a skilled team with in-depth knowledge of distributed systems." | automq.com |
| AWS SQS | The npm package sqs-dashboard exists specifically because the AWS console is too limited. | npmjs.com |
| RabbitMQ | Teams become "accidental RabbitMQ experts" firefighting clustering, memory, and disk issues. | detectify.com |
| Upstash QStash | Real observability is gated behind a $200/mo Prod Pack. No framework integration — write your own worker and adapter. | upstash.com |
| Upstash Redis + BullMQ | A single worker's BRPOPLPUSH loop can burn the 500k-command free tier in minutes. | github.com |
| Hatchet | Free tier is metered from dollar one. Real features (audit, retention >3d, multi-tenant) locked behind $500/mo. | hatchet.run |
| Vercel Queues | Only Next.js-on-Vercel. If your worker is NestJS, Laravel, Django, or anywhere else — no support. | vercel.com |
| AsyncBase | ~60s signup → first message. Framework-native SDKs (NestJS, Laravel, Node, Python, Go). First-party MCP. Live dashboard with DLQ redrive. $0 when idle. | this page |
Every other option makes you wire up Grafana, CloudWatch, or a third-party Bull Board. AsyncBase ships with it. DLQ count. Live tail. API-key rotation. Audit log. In the free tier.
Three plans. Dashboard included. No command-count math. No surprise ops tax.
* Dashboard is included on every tier. You pay for seriousness and scale, not for the right to debug. Billing upgrades ship Q2 2026 — until then email us for paid plans.
Examples: weekend project = $0 · launched indie SaaS = $19 · growing product = $23 · multi-service app = $99+
Start shipping the feature that actually matters.