# Messaging and event streaming Kit: Apache projects for small business

October 30, 2026. Research checked October 29, 2026. [Read the article](https://jwatte.com/blog/apache-messaging-event-streaming-small-business/).

## How to use this kit

Use this with the person who maintains your applications. Write down one failed handoff before choosing a broker. All eleven projects are IT/provider tools in this catalog; there are no owner-run project cards. The examples are illustrations, not installations we performed. Price application changes separately from infrastructure and use sample data during evaluation.

## Provider cards

### ActiveMQ Classic

Purpose: persistent work queues for applications already supporting this broker. Examples: an HVAC company sends completed jobs to invoicing; a print shop holds finishing requests during an outage. Needs IT for Java, persistence, authentication and replay-safe consumers. The current release matrix supports Java starting at 17; the old getting-started disk figures concern 5.x, not a present production sizing guarantee. There is no verified universal RAM minimum; our bounded assumption is one Medium 8 GiB server.

Start with [current downloads](https://activemq.apache.org/components/classic/download/) and [official installation](https://activemq.apache.org/components/classic/documentation/getting-started). Ask: Which version does our application support? Who watches failed jobs? Can we restore messages without sending duplicate invoices? Amazon MQ is the managed broker comparison; match engine versions and application behavior before purchase.

### Artemis

Purpose: message queues for custom integrations using its supported protocols. Examples: a retailer queues packaging requests; a bakery queues wholesale order events. Needs IT for Java 17 or later, separate broker instance configuration, access controls and persistence. No universal RAM minimum is published in the cited guidance. Our allowance is one Small 4 GiB server for a bounded pilot, not an availability promise.

Start with [downloads](https://artemis.apache.org/components/artemis/download/), [broker-instance setup](https://artemis.apache.org/components/artemis/documentation/latest/using-server.html) and [memory/paging guidance](https://artemis.apache.org/components/artemis/documentation/latest/perf-tuning.html). Ask: When is an event acknowledged? Can old messages expire safely? What tells staff that a request is stuck? SQS is an alternative only if the application can use its API and delivery model.

### Kafka

Purpose: retained event history for independent consumers. Examples: a delivery company separates dispatch events from reporting; a two-location parts retailer feeds stock changes to replenishment and its store. Needs IT for storage, replication, permissions, consumer positions and recovery. The official quick start supports Docker or Java 17 or later. The KRaft guide describes three or five controllers, with a typical controller allowance of 5 GB memory and 5 GB disk. Our modeled small cluster uses three controllers and three brokers on six Medium servers; broker count and size need workload testing.

Start with [quick start](https://kafka.apache.org/quickstart/), [KRaft roles](https://kafka.apache.org/44/operations/kraft/), [hardware guidance](https://kafka.apache.org/44/operations/hardware-and-os/) and [releases](https://kafka.apache.org/community/downloads/). Ask: Which application owns inventory truth? How long is replay needed? What is the plan for a bad event schema? Confluent Cloud offers managed Kafka; price connectors, network features and capacity as well as storage.

The monthly hours below are our assumptions: 1 to 3 for a deliberately limited single-server evaluation, or 6 to 16 for a small production cluster. They exclude development and incidents. Where a documented quick start has no numerical hardware floor, we say so; we do not turn an example into a production guarantee.

**BifroMQ (Incubating).** This is MQTT infrastructure for connecting devices and applications. Its [quick-install guide](https://bifromq.incubator.apache.org/docs/get_started/quick_install/) starts one Docker container and verifies publishing and subscriptions. That is the smallest demonstrated setup. Production capacity and a RAM floor are unspecified there. Device credentials, network isolation and reconnect behavior need IT review. Allow 1 to 3 hours/month for evaluation and 6 to 16 if the provider builds a cluster. For example, a two-site car wash could collect equipment status through MQTT. [HiveMQ Cloud Starter](https://www.hivemq.com/pricing/) publishes a managed alternative from $0.34/hour plus $0.80/million messages. At our assumed 730 hours and one million messages, that is $249/month before extras, without an annual term. Protocol compatibility and throughput limits still need checking. Provider questions: How are device credentials rotated? What happens when a site reconnects with a backlog?

**BookKeeper.** This stores replicated logs for other systems; it is not the queue a receptionist uses. Its [local guide](https://bookkeeper.apache.org/docs/getting-started/run-locally/) runs several bookies for learning, and its [deployment guide](https://bookkeeper.apache.org/docs/deployment/manual/) adds metadata coordination and separate bookie storage. A provider must size that stack and its quorum before claiming a production minimum. Budget our cluster range of 6 to 16 hours/month. A regional delivery firm might meet it underneath Pulsar. Confluent Basic’s published $0.05/GB traffic and $0.08/GB-month storage provide an alternative event-service budget. Changing the underlying storage architecture requires a separate assessment. Compare the parent service rather than buying a log subsystem alone. Provider questions: Who operates metadata coordination and the storage quorum? Can we restore the parent service from independent copies?

**EventMesh.** This routes events between applications and messaging backends. The [current quick start](https://eventmesh.apache.org/docs/quickstart/getting-started/) offers a Docker runtime with an in-memory backend explicitly limited to development and testing; a real deployment adds Kafka or RocketMQ. The smallest trial is one runtime, while production inherits the backend’s requirements. Treat current online documentation separately from the table’s 1.12.0 release and confirm version compatibility. A twelve person manufacturer linking an existing order system to warehouse events might justify it. Allow 6 to 16 hours/month for the combined stack. SQS’s one million free request units/month is a priced alternative for a simpler newly written integration, with application changes required. Provider questions: Which released runtime and backend versions work together? How will failed deliveries be reconciled?

**Iggy.** This supplies a streaming server and client libraries. Its [getting-started guide](https://iggy.apache.org/docs/introduction/getting-started/) uses an official 0.9.0 Docker image and requires Linux kernel 5.19 or later. The smallest example is one server; no business workload RAM floor is established there. Container permissions, persistence and client behavior need a developer and operator. A ten person print shop with a custom status feed might evaluate it, allowing 1 to 3 administration hours/month for the pilot. Confluent Basic’s usage rates above are the managed streaming comparison; application and protocol differences prevent a simple server swap. Provider questions: Which container permissions are actually required? How is persisted data recovered after host loss?

**Pulsar.** A broker and storage system work together to deliver messages and streams. The [5.0 standalone guide](https://pulsar.apache.org/docs/5.0.x/getting-started-standalone/) runs a learning environment in one process, and requires a 64-bit Java 21 or later runtime for servers. That is not a production availability design or a universal RAM minimum. A provider needs to scope brokers, BookKeeper storage and coordination separately; assume 6 to 16 administration hours/month for a small cluster. A two-location ecommerce business with several independent consumers might justify a review. Confluent Basic is a priced event-stream alternative using the rates above, subject to a client migration assessment because its protocols differ. Provider questions: How many brokers, bookies and coordination nodes are included in the quote? Who tests a failure of each role?

**Qpid.** The catalog’s release refers to Broker-J; other Qpid components have separate versions. Its [installation documentation](https://qpid.apache.org/releases/qpid-broker-j-10.0.1/book/Java-Broker-Installation.html) describes a Java broker installed from an archive. Use the matching current manual before selecting the runtime; a one-broker trial does not establish failover capacity. A distributor whose vendor already requires AMQP might use it. Allow 1 to 3 hours/month for a bounded single server, more for a cluster. Amazon MQ’s ActiveMQ price example above provides a published broker-service comparison, but the vendor must approve protocol features, authentication and application behavior before substitution. Provider questions: Which Qpid component and Java release does our vendor support? Has a managed broker passed application compatibility tests?

**RocketMQ.** Its [local quick start](https://rocketmq.apache.org/docs/quickStart/01quickstart/) starts a NameServer, broker and proxy, then sends and receives test messages. That identifies a small development topology. Production memory still needs sizing. The operator must size storage and plan recovery across those roles. A growing online retailer with software already built for RocketMQ might justify it; budget 6 to 16 administration hours/month for a small production cluster. SQS’s published free request allowance is an alternative for a simpler new queue workload, with a separate rewrite estimate. Do not price only the broker and forget the applications. Provider questions: Who owns the NameServer, proxy and broker updates? What is the tested recovery sequence?

**Vysper.** This is an XMPP messaging server, a different job from Kafka. Its [user guide](https://mina.apache.org/vysper-project/user_guide.html) is sparse and the catalog’s 0.7 archive dates to 2012. A current supported Java and hardware baseline was not verified, so we would not start a new business deployment from it. A nonprofit with a legacy XMPP integration should have its provider inventory the dependency and price migration. Allow at least our 1 to 3 hours/month planning range for a legacy single server, without pretending that buys security support. [Slack Pro](https://slack.com/pricing) lists $8.75 per person/month billed monthly, or $7.25 with annual billing, before temporary promotions. It can replace the staff-chat job, with any XMPP integrations separately migrated. Provider questions: Can anyone commit to supporting this old release and its dependencies? What would migration to maintained chat software require?


## Backup, update and first-week checks

- Record the application’s supported broker version, event format and contact responsible for failed work.
- Keep configuration and retained message data recoverable through a documented backup process appropriate to the selected broker. Protect credentials separately; an untested copy of a running data directory is not proof of a usable backup.
- Agree on how much data may be lost and how long a recovery may take. Test restoration in an isolated environment and reconcile business identifiers afterward.
- Read security advisories before updating, test the application against the new broker and document rollback limitations for stored data.
- Day 1: identify a single handoff and collect sample events. Day 2: let the provider create a private trial. Day 3: interrupt a consumer and observe the backlog. Day 4: replay events and check for duplicate effects. Day 5: restore from backup and have staff identify failed work in the proposed screen. Stop if nobody owns reconciliation.

For Kafka, also document topic retention and consumer offsets. For Classic and Artemis, document persistence settings, acknowledgement rules and the treatment of repeatedly failed messages. For every choice, keep a manual business procedure available during the test.

## Full group table

Incubating means not yet fully endorsed by the ASF; see the [official explanation](https://incubator.apache.org/guides/branding.html). Archive dates are file timestamps. ActiveMQ and Kafka dates below use freshly checked publisher release dates.

| Project | What it does | Who runs it | Latest release and date |
|---|---|---|---|
| [Apache ActiveMQ](http://activemq.apache.org/) | Queues messages so business applications can exchange work reliably. | Your IT department or provider | [6.4.0; 2026-10-27](https://activemq.apache.org/components/classic/download/) |
| [Apache Artemis](https://artemis.apache.org) | Queues and routes messages between business applications. | Your IT department or provider | [2.57.0; 2026-09-09](https://artemis.apache.org/components/artemis/download/) |
| [Apache BifroMQ (Incubating)](https://bifromq.incubator.apache.org/) | Connects many devices and applications through a shared MQTT messaging service. | Your IT department or provider | [4.0.0-incubating; 2026-01-28](https://github.com/apache/bifromq/releases/tag/v4.0.0-incubating) |
| [Apache BookKeeper](http://bookkeeper.apache.org/) | Keeps replicated event logs that other distributed systems depend on. | Your IT department or provider | [4.18.1; 2026-09-16](https://github.com/apache/bookkeeper/releases/tag/release-4.18.1) |
| [Apache EventMesh](https://eventmesh.apache.org) | Routes events between applications and messaging systems. | Your IT department or provider | [1.12.0; 2025-12-29](https://eventmesh.apache.org/download/) |
| [Apache Iggy](https://iggy.apache.org) | Stores and delivers message streams for applications. | Your IT department or provider | [0.9.0; 2026-09-18](https://iggy.apache.org/downloads/) |
| [Apache Kafka](https://kafka.apache.org/) | Carries streams of business events between applications and data systems. | Your IT department or provider | [4.4.0; 2026-10-22](https://kafka.apache.org/community/downloads/) |
| [Apache Pulsar](https://pulsar.apache.org/) | Stores and delivers messages and event streams across applications. | Your IT department or provider | [5.0.1; 2026-10-26](https://pulsar.apache.org/download/) |
| [Apache Qpid](http://qpid.apache.org) | Provides messaging servers and client tools using standard messaging protocols. | Your IT department or provider | [Broker-J 10.1.1; 2026-09-20 (archive date)](https://downloads.apache.org/qpid/broker-j/10.1.1/) |
| [Apache Rocketmq](https://rocketmq.apache.org) | Delivers messages and events between distributed applications. | Your IT department or provider | [5.5.1; 2026-08-20](https://github.com/apache/rocketmq/releases/tag/rocketmq-all-5.5.1) |
| [Apache Vysper](http://mina.apache.org/vysper-project) | Hosts instant messaging services using the XMPP protocol. | Your IT department or provider | [0.7; 2012-12-26 (archive date)](https://downloads.apache.org/mina/vysper/0.7/) |

## Cost worksheet

Business and application: ______. Required protocol: ______. Maximum tolerated interruption: ______.

Setup hours low/high: ______ / ______. Administration hours/month low/high: ______ / ______. Nodes by size: ______. Retained backup GB low/high: ______ / ______. Managed plan, region, billing term and price-page date: ______. Managed setup cost: ______. Application development and migration quote for each route: ______.

The shared method uses $68.09 per administration hour, an employee compensation estimate from [BLS wages](https://data.bls.gov/oesprofile/?areas=INDUSTRY,STATE,MSA&major_group=150000&measure=01&occupation=151244&year=2025) divided by the [70.0% wages share](https://data.bls.gov/timeseries/CMU2020000000000P). An outside IT provider bills more than this proxy, so get a quote. [DigitalOcean](https://www.digitalocean.com/pricing/droplets) supplies our Small server at $24/month (2 vCPU, 4 GiB RAM, 80 GiB disk), Medium at $48 (4 vCPU, 8 GiB, 160 GiB), and Large at $252 (8 dedicated vCPU, 32 GiB, 100 GiB). These are monthly caps without an annual commitment. [Backblaze B2](https://www.backblaze.com/cloud-storage/pricing) costs $6.95 per decimal TB/month after the first 10 GB free. All were checked October 29, 2026.

Setup = hours × $68.09. Monthly self hosted = server charges + max(0, retained backup GB minus 10) × $0.00695 + administration hours × $68.09. The 36 month total adds setup to 36 monthly bills; savings subtract that total from the alternative. Break even is setup divided by positive monthly savings, rounded up; otherwise it is never. Calculations retain cents internally and display whole dollars.

Our assumptions are 4 to 12 setup hours and 1 to 3 administration hours/month for a bounded single server, or 24 to 60 and 6 to 16 for a small cluster. These exclude application development, migration, training, incidents and on call coverage on both sides. Alternative setup is assumed zero incremental infrastructure labor in these examples. Implementation still takes work. Backup volumes include retained copies. We assume no chargeable backup requests or excess transfer; add them if your usage exceeds the provider allowances. Low and high cases are planning ranges to replace with measured requirements.

### Filled example: ActiveMQ

Our assumptions: one $48 Medium server, 50 to 100 GB backups, setup 4 to 12 hours, administration 1 to 3 hours/month. [Amazon MQ](https://aws.amazon.com/amazon-mq/pricing/) publishes $0.288/hour per mq.m5.large ActiveMQ broker and $0.10/GB-month EBS in its US East (N. Virginia) example. One broker × 730 hours × $0.288 + 200 GB × $0.10 = $230.24/month, excluding networking. Monthly self hosted low = $48 + 40 × $0.00695 + $68.09 = $116.368; high = $48 + 90 × $0.00695 + 3 × $68.09 = $252.8955.

| ActiveMQ single server | Low labor/workload | High labor/workload |
|---|---:|---:|
| Setup labor | $272 | $817 |
| Monthly self hosted | $116 | $253 |
| Self hosted, 36 months | $4,462 | $9,921 |
| Amazon MQ, monthly | $230 | $230 |
| Amazon MQ, 36 months | $8,289 | $8,289 |
| Self hosting savings, 36 months | $3,827 | Negative $1,633 |
| Break even month | 3 | Never |

### Filled example: Artemis

Our assumptions: one $24 Small server, 10 to 50 GB backups, setup 4 to 12 hours, administration 1 to 3 hours/month. [SQS](https://aws.amazon.com/sqs/pricing/) includes one million requests free each month for all customers. Our 100,000 to 500,000 request units fit that allowance, so modeled queue-service fees are zero. Count polling, sends, receives, deletes and size units; other AWS costs and application conversion remain extra. Self hosted monthly low = $24 + $0 + $68.09 = $92.09; high = $24 + 40 × $0.00695 + 3 × $68.09 = $228.548.

| Artemis single server | Low labor/workload | High labor/workload |
|---|---:|---:|
| Setup labor | $272 | $817 |
| Monthly self hosted | $92 | $229 |
| Self hosted, 36 months | $3,588 | $9,045 |
| SQS request allowance, monthly | $0 | $0 |
| SQS request allowance, 36 months | $0 | $0 |
| Self hosting savings, 36 months | Negative $3,588 | Negative $9,045 |
| Break even month | Never | Never |

### Filled example: Kafka

Our assumptions: six $48 Medium servers; 100 to 500 GB backups; setup 24 to 60 hours; administration 6 to 16 hours/month. [Confluent Basic](https://www.confluent.io/pricing/) lists first eCKU free, then $0.14/eCKU-hour, $0.05/GB in/out and $0.08/GB-month stored. Assume usage stays within that first compute unit, with 10 to 100 GB each of ingress, egress and average storage. Managed monthly = (10 + 10) × $0.05 + 10 × $0.08 = $1.80 low; (100 + 100) × $0.05 + 100 × $0.08 = $18 high. Extra compute, connectors and network features need a separate quote. Self hosted low = 6 × $48 + 90 × $0.00695 + 6 × $68.09; high = 6 × $48 + 490 × $0.00695 + 16 × $68.09.

| Kafka six servers | Low labor/workload | High labor/workload |
|---|---:|---:|
| Setup labor | $1,634 | $4,085 |
| Monthly self hosted | $697 | $1,381 |
| Self hosted, 36 months | $26,732 | $53,796 |
| Confluent Basic usage, monthly | $2 | $18 |
| Confluent Basic usage, 36 months | $65 | $648 |
| Self hosting savings, 36 months | Negative $26,667 | Negative $53,148 |
| Break even month | Never | Never |

These are infrastructure comparisons with different service features, not certified equivalent systems. The managed option still needs someone to operate the application. All vendor prices were read October 29, 2026, in US dollars before tax. No annual discount or temporary trial credit is used in the filled examples.

### Setup time when Claude or ChatGPT does the typing

October 30, 2026 update. See the [post’s AI setup comparison](https://jwatte.com/blog/apache-messaging-event-streaming-small-business/#setup-time-when-claude-or-chatgpt-does-the-typing).

We keep the ActiveMQ Classic example: one $48 Medium server and 50 to 100 GB of backups. This remains a provider job even with single-server hours. All hours are our planning assumptions, not measured installations. Human review is already included: 1 to 3 setup hours and 0.5 to 1.5 hours/month. Low and high are paired scenarios. We keep hosting and backups unchanged, with monthly H = $48 + max(0, retained GB minus 10) × $0.00695. Existing exclusions still apply.

| ActiveMQ Classic: paired planning cases | Low | High |
|---|---:|---:|
| Manual setup hours | 4 | 12 |
| AI setup share; hours | 50%; 2 | 80%; 9.6 |
| Manual administration hours/month | 1 | 3 |
| AI administration share; hours/month | 75%; 0.75 | 100%; 3 |
| Manual total, 36 months | $4,462 | $9,921 |
| Claude Pro subscription total, 36 months | $3,913 | $9,958 |
| ChatGPT Plus subscription total, 36 months | $3,913 | $9,958 |
| API pay per token: Claude Opus 5.5 total, 36 months | $3,727 | $9,872 |
| API pay per token: GPT-6 Astra total, 36 months | $3,750 | $10,043 |
| Opus setup/monthly token charges | $0.764 / $0.382 | $6 / $3 |
| Astra setup/monthly token charges | $1.970 / $0.985 | $15 / $7.50 |

Let S/A be manual setup/monthly hours, q/r the AI shares, H monthly hosting plus backups, and T0/Tm setup/monthly tool charges. We use $68.09/hour and round totals only after calculation.

```text
Manual total = S × $68.09 + 36 × (H + A × $68.09)
AI total = S × q × $68.09 + T0 + 36 × (H + A × r × $68.09 + Tm)
Savings = Manual total minus AI total
```

AI setup = S × q × $68.09 + T0. AI monthly = H + A × r × $68.09 + Tm. Substitute your own measured hours before deciding.

Claude Pro and ChatGPT Plus each cost $20/month, billed monthly. We charge one full separate setup month, then 25% ($5) for each of 36 operating months only if other work uses the remaining allowance. Without sharing, charge the full price and add $540 to either subscription total. Never allocate more than 100% across projects. Equal prices do not imply equal usage limits or results; extra usage and repeated setup months cost more. API billing is separate, with no assumed subscription credits.

Our cumulative API setup budget is 200,000 to 1,000,000 input tokens and 20,000 to 100,000 output tokens; monthly administration is 100,000 to 500,000 input and 10,000 to 50,000 output. These are our assumptions including retries, repeated context and billed reasoning, not a single prompt. Keep each request in the applicable short context pricing band. Low input is 30% ordinary, 10% short cache writes and 60% cache reads; high input is 100% ordinary with caching disabled. Do not assume a cache survives between monthly sessions.

Standard API rates per million tokens, read October 29, 2026: Claude Opus 5.5 input $4, output $20, cache read $0.20, short cache write $5; GPT-6 Astra input $10, output $50, cache read $1, short cache write $12.50. Writes replace ordinary input charges. No batch discount, paid browser service, premium routing or storage charge is assumed.

```text
API tool cost = (ordinary input × input rate + cache writes × write rate
                + cache reads × read rate + output × output rate) / 1,000,000
Opus low setup = 0.06 × $4 + 0.02 × $5 + 0.12 × $0.20 + 0.02 × $20 = $0.764
Astra low setup = 0.06 × $10 + 0.02 × $12.50 + 0.12 × $1 + 0.02 × $50 = $1.970
```

The base uses free local Desktop Commander or Codex CLI. Optional Desktop Commander Remote Pro is $20/month; one setup month plus a justified 25% monthly share adds $200 over this period. With exclusive use, its full monthly fee applies. Model usage and hosting are separate.

The high case saves no monthly labor. Compared with manual work, shared subscriptions become more expensive in month 29, Opus API in month 53 and Astra API in month 20. For a slower-work sensitivity, use 119% of manual hours plus tool charges; that is a sensitivity inspired by a coding study, not a deployment forecast. [66] The conclusion stays: the low case beats Amazon MQ’s $8,289 over 36 months; the high case loses. AI typing does not supply broker failover.

Use a staging or new server; never paste secrets into chat. Approve commands or use a restricted sandbox, and watch browser actions. Use supported subscriptions under human direction and API keys for custom automation. Review changes and test a queue restore and duplicate-message handling; backups and updates remain our responsibility.

## Security and licensing

Check [ActiveMQ security](https://activemq.apache.org/security-advisories), [Artemis security](https://artemis.apache.org/security-advisories) and [ASF security guidance](https://www.apache.org/security/) before approving a version. Check release activity and the [Attic](https://attic.apache.org/) rather than assuming directory membership means current support. [Apache License 2.0](https://www.apache.org/licenses/LICENSE-2.0) allows commercial use subject to its conditions and supplies no warranty or trademark license.

## Fact-check notes and sources

Official pages and the series catalog evidence were read October 29, 2026. Prices are US dollars before tax; billing terms appear with each comparison. We have not installed these projects for this series.

1. [current downloads](https://activemq.apache.org/components/classic/download/). Read October 29, 2026, 21:39:20 UTC.
2. [official installation](https://activemq.apache.org/components/classic/documentation/getting-started). Read October 29, 2026, 21:37:52 UTC.
3. [downloads](https://artemis.apache.org/components/artemis/download/). Read October 29, 2026, 21:37:53 UTC.
4. [broker-instance setup](https://artemis.apache.org/components/artemis/documentation/latest/using-server.html). Read October 29, 2026, 21:37:52 UTC.
5. [memory/paging guidance](https://artemis.apache.org/components/artemis/documentation/latest/perf-tuning.html). Read October 29, 2026, 21:38:46 UTC.
6. [quick start](https://kafka.apache.org/quickstart/). Read October 29, 2026, 21:37:52 UTC.
7. [KRaft roles](https://kafka.apache.org/44/operations/kraft/). Read October 29, 2026, 21:38:47 UTC.
8. [hardware guidance](https://kafka.apache.org/44/operations/hardware-and-os/). Read October 29, 2026, 21:40:16 UTC.
9. [releases](https://kafka.apache.org/community/downloads/). Read October 29, 2026, 21:39:18 UTC.
10. [quick-install guide](https://bifromq.incubator.apache.org/docs/get_started/quick_install/). Read October 29, 2026, 21:40:16 UTC.
11. [HiveMQ Cloud Starter](https://www.hivemq.com/pricing/). Read October 29, 2026, 21:37:58 UTC.
12. [local guide](https://bookkeeper.apache.org/docs/getting-started/run-locally/). Read October 29, 2026, 21:40:17 UTC.
13. [deployment guide](https://bookkeeper.apache.org/docs/deployment/manual/). Read October 29, 2026, 21:39:19 UTC.
14. [current quick start](https://eventmesh.apache.org/docs/quickstart/getting-started/). Read October 29, 2026, 21:39:18 UTC.
15. [getting-started guide](https://iggy.apache.org/docs/introduction/getting-started/). Read October 29, 2026, 21:37:57 UTC.
16. [5.0 standalone guide](https://pulsar.apache.org/docs/5.0.x/getting-started-standalone/). Read October 29, 2026, 21:39:20 UTC.
17. [installation documentation](https://qpid.apache.org/releases/qpid-broker-j-10.0.1/book/Java-Broker-Installation.html). Read October 29, 2026, 21:40:30 UTC.
18. [local quick start](https://rocketmq.apache.org/docs/quickStart/01quickstart/). Read October 29, 2026, 21:37:57 UTC.
19. [user guide](https://mina.apache.org/vysper-project/user_guide.html). Read October 29, 2026, 21:40:30 UTC.
20. [Slack Pro](https://slack.com/pricing). Read October 29, 2026, 21:37:59 UTC.
21. [official explanation](https://incubator.apache.org/guides/branding.html). Read October 29, 2026, 21:37:59 UTC.
22. [Apache ActiveMQ](http://activemq.apache.org/). Read October 29, 2026, 21:12:49 UTC (catalog observation).
23. [Apache Artemis](https://artemis.apache.org). Read October 29, 2026, 21:12:52 UTC (catalog observation).
24. [Apache BifroMQ (Incubating)](https://bifromq.incubator.apache.org/). Read October 29, 2026, 21:38:47 UTC.
25. [4.0.0-incubating; 2026-01-28](https://github.com/apache/bifromq/releases/tag/v4.0.0-incubating). Read October 29, 2026, 21:19:11 UTC (catalog observation).
26. [Apache BookKeeper](http://bookkeeper.apache.org/). Read October 29, 2026, 21:38:46 UTC.
27. [4.18.1; 2026-09-16](https://github.com/apache/bookkeeper/releases/tag/release-4.18.1). Read October 29, 2026, 21:22:49 UTC (catalog observation).
28. [Apache EventMesh](https://eventmesh.apache.org). Read October 29, 2026, 21:38:47 UTC.
29. [1.12.0; 2025-12-29](https://eventmesh.apache.org/download/). Read October 29, 2026, 21:13:25 UTC (catalog observation).
30. [Apache Iggy](https://iggy.apache.org). Read October 29, 2026, 21:13:38 UTC (catalog observation).
31. [0.9.0; 2026-09-18](https://iggy.apache.org/downloads/). Read October 29, 2026, 21:14:12 UTC (catalog observation).
32. [Apache Kafka](https://kafka.apache.org/). Read October 29, 2026, 21:13:39 UTC (catalog observation).
33. [Apache Pulsar](https://pulsar.apache.org/). Read October 29, 2026, 21:13:44 UTC (catalog observation).
34. [5.0.1; 2026-10-26](https://pulsar.apache.org/download/). Read October 29, 2026, 21:15:28 UTC (catalog observation).
35. [Apache Qpid](http://qpid.apache.org). Read October 29, 2026, 21:13:44 UTC (catalog observation).
36. [Broker-J 10.1.1; 2026-09-20 (archive date)](https://downloads.apache.org/qpid/broker-j/10.1.1/). Read October 29, 2026, 21:25:11 UTC (catalog observation).
37. [Apache Rocketmq](https://rocketmq.apache.org). Read October 29, 2026, 21:13:45 UTC (catalog observation).
38. [5.5.1; 2026-08-20](https://github.com/apache/rocketmq/releases/tag/rocketmq-all-5.5.1). Read October 29, 2026, 21:16:03 UTC (catalog observation).
39. [Apache Vysper](http://mina.apache.org/vysper-project). Read October 29, 2026, 21:38:48 UTC.
40. [0.7; 2012-12-26 (archive date)](https://downloads.apache.org/mina/vysper/0.7/). Read October 29, 2026, 21:25:11 UTC (catalog observation).
41. [BLS wages](https://data.bls.gov/oesprofile/?areas=INDUSTRY,STATE,MSA&major_group=150000&measure=01&occupation=151244&year=2025). Read October 29, 2026, 21:12:08 UTC.
42. [70.0% wages share](https://data.bls.gov/timeseries/CMU2020000000000P). Read October 29, 2026, 21:17:59 UTC.
43. [DigitalOcean](https://www.digitalocean.com/pricing/droplets). Read October 29, 2026, 21:10:42 UTC.
44. [Backblaze B2](https://www.backblaze.com/cloud-storage/pricing). Read October 29, 2026, 21:10:42 UTC.
45. [Amazon MQ](https://aws.amazon.com/amazon-mq/pricing/). Read October 29, 2026, 21:38:33 UTC.
46. [SQS](https://aws.amazon.com/sqs/pricing/). Read October 29, 2026, 21:38:53 UTC.
47. [Confluent Basic](https://www.confluent.io/pricing/). Read October 29, 2026, 21:38:42 UTC.
48. [ActiveMQ security](https://activemq.apache.org/security-advisories). Read October 29, 2026, 21:37:54 UTC.
49. [Artemis security](https://artemis.apache.org/security-advisories). Read October 29, 2026, 21:39:20 UTC.
50. [ASF security guidance](https://www.apache.org/security/). Read October 29, 2026, 21:39:57 UTC.
51. [Attic](https://attic.apache.org/). Read October 29, 2026, 22:09:30 UTC.
52. [Apache License 2.0](https://www.apache.org/licenses/LICENSE-2.0). Read October 29, 2026, 21:39:56 UTC.

53. [Claude plans](https://claude.com/pricing). Read October 29, 2026, 23:31:08 UTC; reused for this update.
54. [ChatGPT plans](https://chatgpt.com/pricing/). Read October 29, 2026, about 23:27 UTC in a regular browser.
55. [Codex plan access](https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan). Read October 29, 2026, 23:31:16 UTC; reused for this update.
56. [Codex CLI](https://developers.openai.com/codex/cli). Read October 29, 2026, 23:32:01 UTC; reused for this update.
57. [Desktop Commander pricing](https://desktopcommander.app/pricing/). Read October 29, 2026, 23:31:08 UTC; reused for this update.
58. [Desktop Commander setup and permissions](https://raw.githubusercontent.com/wonderwhy-er/DesktopCommanderMCP/main/README.md). Read October 29, 2026, 23:31:41 UTC; reused for this update.
59. [Claude in Chrome](https://support.claude.com/en/articles/12012173-get-started-with-claude-in-chrome). Read October 29, 2026, 23:31:11 UTC; reused for this update.
60. [Chrome permissions](https://support.claude.com/en/articles/12902446-claude-in-chrome-permissions-guide). Read October 29, 2026, 23:31:09 UTC; reused for this update.
61. [ChatGPT cloud browser](https://help.openai.com/en/articles/20001280-using-cloud-browser-in-chatgpt). Read October 29, 2026, 23:32:01 UTC; reused for this update.
62. [Claude API pricing](https://platform.claude.com/docs/en/about-claude/pricing). Read October 29, 2026, 23:31:14 UTC; reused for this update.
63. [OpenAI API pricing](https://developers.openai.com/api/docs/pricing). Read October 29, 2026, 23:31:13 UTC; reused for this update.
64. [OpenAI prompt caching](https://developers.openai.com/api/docs/guides/prompt-caching). Read October 29, 2026, 23:32:00 UTC; reused for this update.
65. [METR coding study limitations](https://metr.org/blog/2026-02-24-uplift-update/). Read October 29, 2026, 23:31:09 UTC; reused for this update.

66. [METR July 2025 developer experiment](https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/). Read October 29, 2026, 23:30:58 UTC; reused for this update.

*Apache, the Apache feather logo and the Apache project names are trademarks of The Apache Software Foundation. This site is not affiliated with or endorsed by the ASF. Other product names are used for identification only.*