Overview

pg-boss is a Node.js job queue that stores and coordinates background work in PostgreSQL. It is best suited to teams already operating PostgreSQL and comfortable with SQL and database-backed operations. Its central advantage is tying jobs to application transactions; at-least-once delivery and a shared job table demand care.

Key features

Workers use PostgreSQL SKIP LOCKED to claim jobs concurrently without blocking one another, with atomic commit safety. More importantly for application consistency, a job can be created in the same transaction as its related data change: a rollback removes both, avoiding queued work for an uncommitted change.

Delivery is at least once, so a handler can run more than once for a job. Teams need handlers that safely tolerate repeat execution. Queue controls include retries with exponential backoff, rate limiting, debouncing, concurrency policies, priorities, and dead-letter queues. These help manage bursts and route exhausted work, but do not remove the need to design for duplicate delivery.

Scheduling supports cron and RRULE patterns, deferred jobs, and dependency-based workflows. Workers use backpressure-aware polling, with optional PostgreSQL LISTEN/NOTIFY for millisecond delivery latency. OpenTelemetry traces and metrics include continuation of a job trace from send() through worker processing, useful for following work across those stages.

Transaction adapters cover Knex, Kysely, Drizzle, and Prisma. Supported database backends include PostgreSQL, CockroachDB, YugabyteDB, Citus, and embedded PGlite. The @pg-boss/dashboard package adds a web interface for monitoring and managing jobs, queues, and schedules. The @pg-boss/proxy package exposes methods through a JSON HTTP API for non-Node runtimes, connection pooling, or scalability; its optional basic authentication, configurable CORS, and route allowlists or denylists provide access controls.

Pricing

pg-boss costs 0.00 USD per free. The MIT-licensed open-source software runs in your own database and has no per-job charges, making it a straightforward fit for teams prepared to operate the required runtime and database. There is no free trial because the plan is free, and no paid plan is stated.

Platforms

pg-boss is self-hosted and supports API, Linux, and web use. It requires Node 22.12 or higher or Bun, plus PostgreSQL 13 or higher. Official SDK languages are JavaScript and TypeScript. Its database requirement makes it a poor fit for teams unwilling to operate PostgreSQL or needing a queue independent of their database stack.

Who it's for

The strongest fit is a team already running PostgreSQL that wants background processing without adding a separate system to operate and monitor. Relational-database fluency matters: users should be comfortable with SQL, querying, and backups. A single-person maintainer means sponsorship supports maintenance, security response, and support; teams evaluating long-term operational continuity should weigh that alongside the MIT license.

The default shared partitioned job table is an important scaling consideration. A queue that grows substantially or develops an unexpected backlog can affect performance through the shared table, so pg-boss is less compelling when queue growth is hard to predict or isolation between queues is essential.

Pros and cons

Pro: Jobs can commit atomically with application data, reducing mismatches between database changes and queued work.
Pro: Scheduling, workflow dependencies, retry policies, rate controls, dead-letter queues, and observability cover a broad range of background-work needs without adding another operated system.
Pro: MIT licensing and no per-job charges suit teams comfortable self-hosting the database-backed service.
Con: At-least-once delivery requires repeat-safe handlers; it is not a fit for work that cannot tolerate duplicate execution.
Con: Queues share a partitioned job table by default, so a large or unexpectedly backed-up queue can affect performance.
Con: Node 22.12 or Bun and PostgreSQL 13 or higher narrow the fit for teams with different runtime or database standards.
Con: Maintenance is handled by one person, a consideration for teams that require broader maintainer coverage.

Alternatives

Choose Sidekiq if you want a freemium queue with a free LGPL open-source option for hobbyists and a commercial Enterprise plan starting at 269.00 USD per month. Celery is another free, open-source distributed task queue, available through PyPI or source, for teams seeking that option across Linux, macOS, and self-hosted environments.

Inngest may suit teams looking for a managed-style free tier with stated seat, execution, event-ingestion, and trace-history allowances, or a Pro plan at 99.00 USD per month. Consider IronWorker if its free Basic plan's public cluster, one concurrency, five hours per month, and 512 MB RAM match your workload; it also offers a free trial.

Trigger.dev offers a free plan with $5 monthly credits, 20 concurrent runs, unlimited tasks, and five team members, which may suit teams comparing execution limits and team capacity. Faktory is an alternative with an AGPL open-source edition and enterprise connection packs starting at 99.00 USD per month.

Pick River if its Open Source plan's web UI and absence of workflows or concurrency limits better match your needs; its Pro plan is 125.00 USD per month. Dramatiq is another free option for Linux and self-hosted use, including Windows.

For a wider comparison, browse Task Queue Software.

Verdict

Choose pg-boss if your team already runs PostgreSQL and wants transactional background jobs, scheduling, and queue controls without operating a separate queue system. Its database fit and lack of per-job charges are compelling; look elsewhere if duplicate delivery is unacceptable, the shared-table behavior is a concern, or your stack cannot meet its runtime and database requirements.

	Oban	pg-boss
Free plan	No	Yes
Free trial	No	No
Paid from	—	Free
Open source	No	No
Platforms	—	api, Linux, self-hosted, Web
Free plan	Yes	Yes
Paid from	135 /mo	—
Deployment model	self_hosted	self_hosted
Custom retry policy	Yes	Yes
Scheduled jobs	Yes	Yes
Official SDK languages	Elixir, Python	JavaScript, TypeScript
Dead-letter queue	—	Yes

Oban vs pg-boss

Oban

pg-boss