Bunqueue
Bunqueue is a free, MIT-licensed job queue for Bun that stores jobs in SQLite or PostgreSQL — no Redis required.
If Redis is the only reason your queue needs a babysitter, Bunqueue deletes that reason: an MIT-licensed server, SQLite or PostgreSQL persistence, one TCP port, and matching clients for TypeScript, Python, PHP, Go, Rust and Elixir. The MCP server — 75 tools in 11 toolsets, 1,300–4,600 tokens of context per request with a toolset loaded versus about 16,000 loaded whole — is the genuinely fresh angle; BullMQ gives you nothing comparable. Just don't mistake it for a clustered broker: the design is single-node, not automatic failover.
Verified 1d ago · liveness 77/100 · cite: rightaichoice.com/tools/bunqueue
- Bun, Node.js and Deno teams that want BullMQ-style jobs without Redis
- Multi-language backends sharing one queue across Python, Go, PHP or Elixir workers
- AI agent builders wiring Claude or Cursor to a queue through the native MCP server
- Solo developers and small teams deploying a single Docker container or VPS
- Systems requiring native clustering, automatic failover or multi-region replication
- Projects that lean on publish/subscribe fan-out, which is outside the queue's scope
- Teams standardized on MySQL or another non-SQLite, non-PostgreSQL store for queue state
We scan live Reddit threads, YouTube comments, X posts, G2 reviews and other communities — and hand you an honest verdict in under a minute.
- Honest verdict, not marketing
- Real pros & cons from real users
- Attributed quotes with receipts
3 free scans · no card needed
Skip Bunqueue if you need automatic failover, multi-region replication or pub/sub fan-out across a cluster — the design is single-node, and 2.9.x ships patches nearly weekly if you want a frozen API.
There is no licence cost to weigh: the server and every queue feature ship under MIT with no paid tier, and you run it on hardware you already pay for. The real comparison is operational — the server plus its SQLite or PostgreSQL store against the Redis instance (and its managed-service bill) that BullMQ needs, or against a hosted queue's per-job pricing.
In short
Bunqueue — Bunqueue is a free, MIT-licensed job queue for Bun that stores jobs in SQLite or PostgreSQL — no Redis required. Best for Bun, Node.js and Deno teams that want BullMQ-style jobs without Redis, Multi-language backends sharing one queue across Python, Go, PHP or Elixir workers, AI agent builders wiring Claude or Cursor to a queue through the native MCP server. Free to use.
What's new in Bunqueue
Checked yesterdayAcross the latest 5 updates: 5 changelog entries.
[2.9.16] - 2026-10-15
Latest patch in the 2.9.x line, published the same day as 2.9.15 and closing a fifteen-release run that started with 2.9.0 on 2026-08-28.
[2.9.15] - 2026-10-15
Patch release published 2026-10-15, one of two versions that landed that day in the October maintenance run.
[2.9.14] - 2026-10-14
Mid-October patch, part of a stretch of releases covering 2.9.12 through 2.9.16 across 2026-10-08 to 2026-10-15.
[2.9.11] - 2026-10-05
Closed out a rapid October patch cluster spanning 2.9.7 through 2.9.11, which shipped across 2026-10-02 to 2026-10-05.
[2.9.8] - 2026-10-03
First of three patches published on 2026-10-03, alongside 2.9.9 and 2.9.10.
What people actually say about Bunqueue — is it worth it?
We ran a structured research pass across product reviews, community discussions, and post-purchase forum threads to surface the patterns vendors won't publish themselves. Below: the recurring strengths, the hidden costs people mention most, and the cohort that consistently regrets adopting this tool.
11 mentions across 2 sources (Hacker News, Lemmy) · researched Jul 3, 2026.
Average across the 2 sources that answered — each source counts once, not each post.
- +Zero external dependencies – no Redis, no RabbitMQ, just Bun and SQLite.
- +Sub-millisecond p99 latency and 150K+ ops/sec on a single node.
- +Drop-in BullMQ compatible API makes migration easy for existing BullMQ users.
- +Built-in dead letter queues, cron scheduling, and workflow engine with sagas.
- +MCP server with 73 tools allows AI agents to manage jobs via natural language.
- −Single-server only – no clustering, no high availability currently.
- −Server downtime blocks all queue operations; no failover mechanism.
- −Tied to Bun runtime – no support for Node.js or Deno.
- −Very young project (released Jan 2026) – limited production maturity.
- −Small community – support relies heavily on the solo creator.
- • No hidden costs – it's free and open source, but you pay in operational risk if you need clustering or HA.
Viability Score
How well maintained and how widely used is Bunqueue? Built from what the vendor actually publishes (docs, changelog, tutorials, integrations, pricing), whether the site is live, and how much real users discuss it. How we calculate this
Last calculated: October 2026
How we score →Key Features
- SQLite-backed job queue with no Redis dependency
- PostgreSQL storage for shared multi-broker setups
- Embedded mode: queue and worker inside your Bun process
- Standalone TCP server on port 6789 with native TLS and auth tokens
- HTTP API on port 6790 for monitoring and health checks
- BullMQ-style API: queue.add() and worker.process()
- Priorities, delayed jobs, rate limits and concurrency controls
- Retries with backoff, stall detection and dead letter queue
- FIFO groups for ordered processing within a queue
- Cron scheduling with expressions, intervals and timezones
- FlowProducer parent-child job dependencies
- Durable workflow engine with saga rollback and human approval gates
- Native MCP server with 75 tools in 11 toolsets over stdio or Streamable HTTP
- MCP confirmations, dynamic toolsets and optional decision models
- Job queries, pagination, Prometheus metrics, webhooks and events
About Bunqueue
Bunqueue is a self-hosted background job queue built for polyglot backends. Your application keeps whatever runtime it already uses — Node.js, Deno, Python, PHP, Go, Rust, Elixir or Bun — and talks to the queue server over TCP on port 6789, with an HTTP API on port 6790 for health probes and monitoring. The server runs on Bun, and job state lives in SQLite by default or PostgreSQL when several brokers need one authoritative store. There is no Redis to install and no broker daemon to babysit. Officially supported clients share a Queue/Worker/Flow core, so semantics stay consistent whether a Python worker picks up what a Go producer enqueued. The API is deliberately BullMQ-shaped: queue.add() and worker.process(), priorities, delayed jobs, FIFO groups, retries with backoff, stall detection, rate limits and concurrency controls. A durable workflow engine layers on parent-child flows, saga rollback and human approval gates. The 2.9.x line has moved fast: fifteen releases between 2026-08-28 and 2026-10-15, including 2.9.16 and 2.9.15 both landing on 2026-10-15. Operationally you get job queries and pagination, Prometheus metrics, webhooks, S3 backup, native TLS, an MQTT integration for IoT and edge work, and Docker images in Alpine, Debian, slim and distroless variants for AMD64 and ARM64 — or you can embed the queue and worker inside a single Bun process with no server to operate. For AI agent builders, a native MCP server exposes 75 tools across 11 toolsets over stdio or Streamable HTTP, with opt-in toolsets, destructive-call confirmations and optional decision models. Everything ships under MIT with no paid tier — the vendor page states that explicitly.
Behind the Verdict
Bunqueue's pitch is narrow and honest: take BullMQ's programming model, drop Redis, and let the queue live in SQLite or the PostgreSQL you already run. That is the right trade for most teams under a few hundred thousand jobs a day. Your job state is in a database you already back up, not a second daemon with its own persistence rules and memory ceiling. Job queries and pagination are plain SQL-shaped instead of Redis key archaeology. The server is small enough to run as one container, or to embed inside a Bun app so there is no server at all. The polyglot story is real engineering rather than a wrapper. Official clients for TypeScript, Python, PHP, Go, Rust and Elixir share a Queue/Worker/Flow core, so a Python worker picks up what a Go producer enqueued with the same retry, priority and stall semantics. The protocol is TCP with native TLS and auth tokens, plus an HTTP API on 6790 for health probes. Docker images come in Alpine, Debian, slim and distroless for both AMD64 and ARM64. Where it stands out for AI work is the MCP server. Claude, Cursor or any MCP client gets 75 tools over stdio or Streamable HTTP. The three opt-in controls are the thoughtful part: BUNQUEUE_MCP_TOOLSETS loads only the groups you need, cutting per-request context from roughly 16,000 tokens to 1,300–4,600; BUNQUEUE_MCP_CONFIRM=destructive makes obliterate, drain and purge show their exact impact and wait for your client to confirm, with a decline always winning; and decision models route plain-language requests to the right toolset and double-check destructive calls — able to block a call but never approve one. Every feature works without a model. The durable workflow engine is the second differentiator: parent-child flows, saga rollback, idempotency and human approval gates, documented as product surface rather than something you assemble from primitives. What it is not: a clustered broker. If you need automatic failover, multi-region replication or publish/subscribe fan-out, this is the wrong tool and the vendor does not pretend otherwise. Storage is SQLite or PostgreSQL, so a MySQL-standardised shop gets nothing. And the release cadence is relentless — fifteen versions between 2026-08-28 and 2026-10-15 — which is good for fixes and bad for anyone who wants an API frozen for years. Against BullMQ, the comparison is mostly operational: same shape, one less daemon, plus MCP and workflows you would otherwise write yourself. Against a managed hosted queue, the trade is control and cost against an SLA and a support contract you will not get here.
Researching Bunqueue? Get your full AI stack in 60 seconds.
Free, no signup — tell us your goal and get tools matched to your budget & existing stack.
Real-world workflow fit
Concrete scenarios for the personas Bunqueue actually fits — and what changes day-one when you adopt it.
You replace a BullMQ setup so you can drop the Redis instance. Point the existing queue.add() and worker.process() calls at port 6789, keep SQLite locally and switch to PostgreSQL once a second broker joins.
Outcome: Same job semantics, one fewer daemon to run, and job state that sits in the database you already back up.
A Go producer enqueues image-processing jobs that Python workers pick up, with retries, priorities and stall detection enforced by the server instead of each client.
Outcome: One shared queue across runtimes with consistent retry and priority behaviour, instead of re-implementing coordination per language.
Register the MCP server with BUNQUEUE_MCP_TOOLSETS=dynamic and BUNQUEUE_MCP_CONFIRM=destructive so Claude or Cursor can enqueue and inspect batch jobs, while obliterate and purge wait for your confirmation.
Outcome: Agents drive the queue at 1,300–4,600 tokens of context per request instead of about 16,000, and destructive calls stay in your hands.
Use Cases
- Schedule email digests with cron and retry failed sends with backoff.
- Orchestrate multi-step data pipelines with saga rollback on failure.
- Let an AI agent enqueue and inspect batch jobs through the MCP server.
- Process webhook payloads with rate limiting and priority queues.
- Offload CPU-intensive image processing from HTTP handlers to workers.
- Share one queue between Node.js, Python, Go, PHP, Rust and Elixir workers.
- Back up queue state to S3 and restore it after a container replacement.
- Run jobs from edge and IoT producers that publish over MQTT.
Models Under the Hood
as of 2026-10-15
Limitations
- Every feature ships under MIT, including the server — the vendor page states there is no paid tier.
- Storage is limited to SQLite (the default) or PostgreSQL; there is no Redis backend to fall back on.
- The design is single-node: no automatic failover, no multi-region replication, no publish/subscribe fan-out.
- Runtime targets are Node.js, Deno, Python, PHP, Go, Rust, Elixir and Bun, with the server itself running on Bun.
- AI-agent features are an MCP server exposing 75 tools in 11 toolsets over stdio or Streamable HTTP, with opt-in confirmations and decision models — every feature works without a model.
- Release cadence is very fast: fifteen 2.9.x releases between 2026-08-28 and 2026-10-15.
as of 2026-10-17
Verification history
We have re-verified Bunqueue 10 times since . Each pass re-reads the vendor's own pages and re-checks every listed field against that evidence; passes where nothing had changed are marked as such.
- — re-checked, vendor evidence unchanged
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
- — re-verified summary, description, our verdict, our analysis, pricing model, pricing tiers, features, integrations, who it suits, who should skip it
Showing the 6 most recent of 10 verification passes.
Free to cite with attribution — this page re-verifies continuously.
12-month cost
Project the real annual outlay, including the implied monthly cost when only an annual tier is published.
Vendor list price only. Add-on usage, seat overages, and contract minimums are surfaced under Hidden costs & gotchas.
Plans compared
For each published Bunqueue tier: who it actually fits, and what it adds vs. the previous tier. Cross-reference the cost calculator above for projected annual outlay.
Community (MIT, self-hosted)
$0/mo
Ideal for
Teams of any size that are willing to run the server themselves on a VPS, in Docker or embedded in a Bun app
What this tier adds
Starting and only tier — full server and every queue feature under MIT, with no paid upgrade above it
Where the pricing makes sense
The company stage and team size where Bunqueue's pricing actually pencils out — and where peers do it cheaper.
There is no licence cost to weigh: the server and every queue feature ship under MIT with no paid tier, and you run it on hardware you already pay for. The real comparison is operational — the server plus its SQLite or PostgreSQL store against the Redis instance (and its managed-service bill) that BullMQ needs, or against a hosted queue's per-job pricing.
Setup time & first value
How long it actually takes to get something useful out of Bunqueue — broken out by persona, not the marketing-page minute.
Docker or VPS deploy: one container plus a connection string gets a first job through in under an hour. Embedded in a Bun app: minutes, since there is no server to start. Migrating from BullMQ: an afternoon if your queue.add()/process() calls already match the API — a dedicated migration guide is in the docs.
Switching to or from Bunqueue
How to bring data in from common predecessors and how to get it back out — written for the switcher, not the buyer.
- →From BullMQ: follow the documented migration guide; the API is deliberately BullMQ-shaped, so queue.add() and worker.process() calls carry over with minimal edits.
- →From Redis-backed queues generally: point producers and workers at TCP port 6789 and let SQLite or PostgreSQL hold job state instead of Redis keys.
- ↗To BullMQ: the queue.add()/worker.process() surface maps back directly, and Redis becomes the store again.
- ↗To a hosted managed queue: export job definitions and re-register cron schedules, since the scheduling model is vendor-specific.
Integrations
Resources & Guides
Tutorials & Learning
YouTube returned 6 videos for “Bunqueue”, and we withheld 6: 6 could not be judged, because “Bunqueue” is a single word that other videos use for other things. We are showing none, because we could not prove any of them are about Bunqueue.
Official links
Featured Head-to-Head Comparisons
Bunqueue vs Spider Cloud
Spider Cloud and Bunqueue serve completely different needs. If your priority is extracting web data for AI agents or RAG, Spider Cloud's Rust-powered API, AI extraction, and 1,000+ scraper catalog make it the clear choice. If you need a high-performance job queue for Bun with no external dependencies, MCP support, and sub-millisecond latency, Bunqueue is unmatched. There's no direct competition between them.
Bunqueue vs Temporal Ai
For polyglot teams building mission-critical, long-running workflows with rich durability guarantees, Temporal's open-source platform is the clear choice. If you're a solo developer or a Bun-native team wanting a lightning-fast, zero-infrastructure job queue with native MCP AI agent support, Bunqueue's free, high-performance SQLite-based solution is hard to beat.
Bunqueue vs Presto Voice
If you run a QSR chain and need to automate drive-thru ordering with proven upselling (up to 6% revenue lift), Presto Voice is your tool—but it's enterprise-priced and not for small restaurants. If you're a Bun developer building AI agent workflows or microservices and want a fast, free, zero-dependency job queue, Bunqueue is the clear winner. These tools serve completely different needs, so choose based on whether you're in the restaurant industry or building software.
Popular in Developer Infrastructure
Temporal AI
Durable execution platform that keeps AI agents and long-running workflows alive through crashes, retries, and abandoned sessions.
Frequently Asked Questions
Categories
Used Bunqueue? Help shape our editorial sentiment research.