A message delivery normally consumes 3 operations (write, read, delete), with additional reads for retries and extra operations for payloads over 64 KB. Queue operations are not a 10.000-messages/day guarantee; Workers Free usage has separate limits. Message retention is fixed to 24 hours on Free, unlike configurable Paid.

Each delivery attempt—including retries—counts against the daily quota; one topic publish costs a delivery per subscribed endpoint. Free includes maximum 7-day delay, 3-day logs/DLQ retention and 10 concurrent deliveries; paid plans increase quotas. Do not confuse published jobs with billed delivery attempts.

A retry counts as another step; context.call/context.invoke each cost two billed steps and parallel execution incurs extra steps. Free workflow sleep is limited to 7 days, log/DLQ retention to 3 days; persistent daily quota overruns can cause workflow failures via QStash. Not unlimited serverless compute.

The service is fully free for regular HTTP cron jobs under fair-use requirements. Jobs call publicly reachable URLs; very high scheduling volume or REST API use is rate limited. Your own target endpoint/cloud provider may still incur compute or network costs.

Workflow steps, state storage, Worker invocations and CPU are separate usage dimensions; 10 ms CPU per invocation on Free and the Workers Free daily request cap apply. Default completed/errored workflow state retention is 3 days. Workflow sleeps/waits do not consume CPU, and Workflows billing began August 10, 2026; Paid has separate metered charges.

No card required. A single function run and every step execution each consume one execution (five steps use six executions). Free includes 100K queue depth, 50 concurrent realtime connections and 24-hour trace/log history; function runs can sleep for at most 30 days. Extra compute/sandbox resources may be billed separately.

The $5 Free compute/run credit renews monthly; it is not a one-time trial. Free includes Dev/Prod environments, 10 concurrent Realtime connections and one-day log and query retention. Usage beyond credit may require an upgrade; Hobby is a separately billed $10/month plan with its own included credits.

Free has 30.000 monthly outbound webhook POST attempts across five queues, no card and no paid overage. Retry attempts count toward the cap; backpressure waiting does not. At the limit you must upgrade rather than assuming unlimited delivery; job retention is three days.

Hosted Windmill Cloud Community has an ongoing bounded monthly execution allocation and is distinct from a separately self-hosted Community Edition. Cloud service limits and regions are provider-operated; personal/commercial use and paid extras should be checked against the latest cloud terms.

Development tier; 28-day idle queue time; unused instances may be deleted after two months of no activity.

Runs on your infrastructure; compute, storage, and third-party API costs are separate. Its fair-code license and community-edition feature limits require review.

You operate the worker and data infrastructure. Review community license and feature differences from paid cloud or enterprise plans.

Provide compute, storage, and any worker infrastructure yourself. Enterprise-specific features and managed hosting are separate.

You operate execution workers, database, and hosting. Check community-edition limitations and paid enterprise features before production adoption.

Requests above the free monthly quota incur charges; data transfer and AWS KMS encryption may cost extra. Unused requests do not roll over. AWS signup payment-method requirements vary by account type and location.

AI model steps use additional credits (2/10/20 per model tier); free quota pauses runs until credits refresh. Official pricing page inconsistently states daily vs monthly reset; verify reset timing. Cloud Free is distinct from self-hosted Community Edition; commercial-use terms not explicitly confirmed.

Active Google Cloud billing account and valid payment method required; operations over 1 million/month cost $0.40 per million, 32 KB chunking and retries count, and HTTP target network/compute can incur separate charges.

Requests beyond the free allowance are listed at $20 per 100K; spending cap can limit charges. Card requirement for free signup and commercial production rights are not explicitly confirmed; Developer is positioned for development.

You supply servers, storage, backups, monitoring and upgrades; capacity is infrastructure-dependent. Community security patches and support are limited to currently supported release series; extended lifecycle and 24/7 vendor support require commercial Tanzu RabbitMQ. Distinct from CloudAMQP's managed Little Lemur tier.

Requires own Redis-compatible data store and worker compute (infrastructure may cost money); BullMQ Pro groups, batches, observables and support are paid and excluded.

Requires your own Go application workers and Redis-compatible broker, plus compute, monitoring, backups and operations. At-least-once task execution requires idempotent handlers; not a hosted queue or free Redis instance. Asynq API is pre-1.0 and Redis Cluster compatibility requires review.

Requires your own Node.js worker processes and PostgreSQL database; database, CPU, storage, backups and monitoring are not included. Tasks run at least once, so handlers should be idempotent; Graphile Worker is not a managed queue or free PostgreSQL service.

Requires operating and securing the Temporal cluster, workers and persistence database (for example PostgreSQL or MySQL); high availability, backups, monitoring and infrastructure incur costs. Temporal Cloud is a separate paid service.

Requires self-managed brokers, persistent storage/BookKeeper, metadata services, network security, monitoring and operational expertise. Cloud infrastructure, managed Pulsar and optional connectors can incur charges.

You must operate your own NATS server/cluster and configure JetStream persistence, storage limits, authentication, replication and backups. Infrastructure, bandwidth and managed NATS/Synadia offerings are separate and may cost money.

Requires your own Kafka brokers/KRaft controllers, compute, disks, networking, monitoring, security, upgrades and backups. Managed Kafka, cloud infrastructure and third-party connectors may have separate charges.

Self-host the API/server, workers, database and task infrastructure; hosting, compute, storage and third-party integrations can cost money. Prefect Cloud managed plans and enterprise features are separate; this is not a free managed cloud allowance.

You operate the Airflow scheduler, API/UI, metadata database, workers and authentication; infrastructure, storage, observability and external integrations can cost money. Managed Airflow hosting such as MWAA/Composer is not included.

Run and maintain the Dagster webserver/daemon, code locations, storage and execution infrastructure yourself; compute, database, logging and external services may cost money. Dagster+ managed Cloud features and support are separate paid products.

Self-hosting needs a database, compute and operational maintenance. External execution engines, cloud resources and managed hosting may incur costs.

Requires your own Kubernetes cluster and operational maintenance; compute, container registry, object storage and networking can incur charges. Not a free hosted workflow service.

Requires operating Hatchet server, PostgreSQL, workers and infrastructure. Hatchet Cloud pricing and payment-card requirements are separate; this entry does not include managed cloud runs.

Requires a Java application and supported SQL/NoSQL storage. The ~100 recurring-job figure is a practical performance guideline, not a hosted allowance. Priority queues, batches and advanced workflow features require JobRunr Pro.

Requires Python worker processes plus a message broker such as RabbitMQ or Redis; infrastructure, broker hosting and monitoring are not included. Native Microsoft Windows support is not maintained.

Requires an Elixir application and a supported database. Oban Pro features including durable workflows, global/rate limiting and queue partitioning are paid; infrastructure is not included.

Self-hosting and operational maintenance are required; external APIs, proxies and infrastructure may have separate charges and limits. Huginn is event/agent automation, not a managed queue service.

The queue server requires Bun; run it on your own infrastructure and pay any VPS/database costs. No managed hosting or SLA is included.

You provide hosting, storage and external API accounts. Enterprise-only features are excluded; AGPL-3.0 network-source-sharing obligations apply to modified deployments.

Self-managed database and worker compute are required; hosting costs are not included. MPL-2.0 obligations apply when distributing modified source files; managed support is separate.

Requires your own PostgreSQL 13+ database and Node.js 22.12+ or Bun; database, compute, storage, backups and operations are not included. Jobs are delivered at least once; handlers must be idempotent.

Batch workloads are pay-as-you-go above both monthly resource allowances; requires a valid payment card; other Scaleway resources billed separately.

Billing account required; free allowance is shared across projects and paused jobs count. Each additional job costs $0.10 per 31 days, and target services may bill separately.

AWS account and billing setup required; invocations above free allowance cost $1 per million. Downstream target services, data transfer and related AWS resources may incur separate charges; GovCloud excluded.

Operate your own server, workers, storage and networking; hosting and external APIs cost extra. Paid self-host licensing adds SSO, RBAC and audit features. GPL-3.0 obligations and separate commercial embedding terms may apply.

Requires your own Node.js hosts, storage and operations; infrastructure costs are not included. Maintainers now focus new features on successor xyOps; Cronicle receives primarily maintenance and security fixes.

Self-managed worker processes and optional Redis/PostgreSQL brokers are not hosted for free; production operation, monitoring and infrastructure costs remain your responsibility.

Embedded in an application rather than a hosted scheduler; a database, worker runtime, servers and operations are supplied and paid for separately. Commercial use must comply with Apache-2.0 notices.

You operate workers and a RabbitMQ/Redis broker; compute, storage and managed broker services may cost extra. LGPL license obligations apply to redistribution or modified library versions; commercial support is separate.

Requires your own Java/Docker host, persistent storage, security configuration and monitoring; infrastructure and operations are not free. Not a managed cloud allowance.

You operate NameServer, brokers, storage, networking and monitoring; Java runtime and infrastructure costs are separate. No free managed cloud instance is included.

Requires your own PostgreSQL database, worker processes, compute, monitoring and backups; hosted Postgres or infrastructure charges are separate. The project is seeking additional maintainers.

Run and secure your own Python workers and Redis/PostgreSQL backend; hosting, database, storage and network costs are not included. SAQ is software, not a managed free cloud queue or hosted SLA.

Self-host Python workers and configure a compatible broker and result backend; Redis/NATS/RabbitMQ/Kafka hosting, compute, monitoring and egress are not included. This is free software, not a vendor-hosted free queue or SLA.

Requires a self-operated Redis >=5 or Valkey >=7.2 broker and Python workers; server, compute, persistence, backups, monitoring and networking are not included. This is not a free managed Redis service; paid Tidelift support is separate.

Runs in your Rails application or workers; requires Rails 7.1+/Ruby 3.2+ and a supported SQL database (SQLite, PostgreSQL or MySQL). Application compute, database hosting, storage and operations are not free. Preserve MIT copyright and license notices.

Runs within your Rails process or separately managed workers and requires PostgreSQL; database, application hosting and operations can incur costs. Retain MIT copyright and license notices.

Runs in your Node.js application/worker environment; persistence and job deduplication are application responsibilities, and any infrastructure or external database costs are separate. Retain MIT notices.

You must provide and operate a database and Node.js process; managed database/compute costs are separate. Agenda 6 requires ESM and a separately installed backend package; migrations from v5 have breaking changes.

No free managed broker, compute, storage, networking, backups or SLA is included. You operate nsqd/nsqlookupd, security, monitoring and data durability; at-least-once delivery requires idempotent consumers. Preserve MIT license notices.

Self-operated Ruby workers and Redis are required; compute, Redis hosting, persistence, monitoring and operations can incur costs. Redis distribution/version licensing and third-party integrations are separate. Current upstream docs require Ruby 3.2+ and Redis server 4.0+; no free managed queue is included.

Requires own Java runtime, cluster compute, memory, network, storage/checkpoints and operations. Cloud-managed Flink offerings and external message brokers/connectors may incur charges; not a hosted free stream-processing plan.

Compute, storage, databases and external services are not free infrastructure. Change default Hop Server credentials, protect REST endpoints with authentication, enable TLS and run only trusted workflows, which execute with host privileges.

Requires self-managed Ruby workers plus Redis, Valkey or Dragonfly and separately funded infrastructure. Sidekiq Pro/Enterprise features are paid. LGPLv3 obligations apply when redistributing Sidekiq; check the license of the chosen datastore.

Requires your own Docker/server, SQLite storage and backups; Telegram notifications require a user-managed bot. Browser notifications require the web app to remain open and HTTPS/localhost; no hosted infrastructure is included. Designed for private Tailscale deployments.

Requires user-operated Redis-compatible storage and Go worker processes. Hosting, monitoring, backups and Redis service costs are separate; preserve MIT license notices.

Runs within user-operated Go applications, not a hosted scheduler. Compute, persistence and optional distributed-lock infrastructure are not included; retain MIT copyright/license notices.

Self-managed workers, message brokers and result backends require infrastructure and operations; paid managed broker services are separate. MPL-2.0 obligations apply to distribution of modified covered files.

Official pricing lists 100K monthly actions, but the vendor's Cloud overview still says 50.000; verify the live account allowance before production use. Each function invocation plus completion consumes two actions, with intermediate steps and larger payloads consuming additional actions. Free has no listed additional-action allowance; throughput is rate-limited. Paid tiers can bill overages. Commercial-use terms for the hosted Free plan are not explicitly confirmed.