IBM MQ AI Agents
Ask, and it investigates.
An administrator asks in natural language, for example where are messages building up?
, and the agents gather live MQ state, reason over it and explain what they find, with guidance on next steps.24
Comparison ยท sources checked 23 October 2026
IBM MQ AI Agents and Abendly: an honest comparison.
IBM now ships AI agents as part of IBM MQ Advanced, generally available since 24 March 2026.1 That's a clear sign of how much MQ teams need help. The two products take different approaches: IBM's agents investigate when an administrator asks. Abendly is built to watch continuously and warn before something breaks. Here is where each one fits, with a source for every statement about IBM.
Where we standAvailable now the log analyser and a one-time MQ health check. Design-partner pilot the monitoring agent compared on this page, read-only on one queue manager in your environment. Roadmap general availability at full estate scale. We'd rather you knew that before reading the table.
Two approaches
Different questions, different moments.
Both use AI to help with IBM MQ. The difference is who starts the conversation, and when.
Abendly
It watches, and warns first.
The agent watches queue depth trends, channels, MQ events and error logs around the clock. When something starts to go wrong it raises a finding before anyone asks: the cause, a confidence level, how long until impact, and the MQSC fix in execution order.
Side by side
The details, row by row.
Where IBM's published materials don't cover a point, we say so instead of guessing. If it matters to you, check with IBM.
| Aspect | IBM MQ AI Agents | Abendly |
|---|---|---|
| Availability | Generally available since 24 March 2026.1 | Monitoring agent: Design-partner pilot on one queue manager. Full-estate general availability is on the roadmap. Log analyser and health check: Available now |
| Mode | Conversational. A natural-language interface for diagnosing and troubleshooting MQ, with a Supervisor Agent as the single entry point.12 | Proactive. Watches continuously and raises a finding when a trend or event starts to go wrong. |
| What starts it | An administrator's question, typed in a chat window in the MQ Web Console (IBM MQ 9.4.5 onwards).4 Unattended watching and alerting aren't described in IBM's published materials; check with IBM. | A depth trend, a channel state change, an MQ event or an error-log entry, at any hour, including 3am. |
| Question it answers | Why aren't messages flowing? Diagnosing message build-up, channel failures and dead-lettered messages.1 | What is about to stop messages flowing, and how long do we have? |
| Time to impact | Not described in IBM's published materials; check with IBM. | Quantified runway on every trend finding, for example “hits MAXDEPTH in about 14 min”. |
| Edition requirement | A component of IBM MQ Advanced, including MQ Advanced in Cloud Pak for Integration; your entitlement extends your MQ Advanced or Cloud Pak for Integration entitlement.13 | Independent of MQ edition. It reads queue managers with standard inquire and display authority and MQ event queues, so it doesn't need MQ Advanced. |
| Coverage | IBM lists IBM MQ and MQ Advanced, MQ Appliance, MQ on z/OS and MQ SaaS. Each agent instance is designed to administer up to 20 queue managers.1 | Log Intelligence reads MQ, CICS and DB2 logs. The monitoring agent covers distributed and z/OS in one view, including CF structures and queue-sharing groups. The pilot covers one queue manager. CICS correlation In development DB2 correlation Roadmap |
| Model choice | Uses large language models to work through natural-language requests.4 Which model providers are supported: check with IBM. | Your own provider and key: Anthropic, Azure OpenAI, Amazon Bedrock, IBM watsonx, or a self-hosted model. Identifiers are replaced with stable labels inside your environment before anything is reasoned over. |
| Deployment | Installed with IBM-supplied Helm charts; supported on Red Hat OpenShift at the time of IBM's April 2026 technical write-up.4 | Runs in your environment: your cloud account or VPC, your data centre, or air-gapped with a self-hosted model. |
| Changes to your queue managers | None made for you. Queue-manager access is read-only; it can generate MQSC commands for an administrator to run.456 | None made for you. Inquire and display authority only, enforced by OAM on distributed and RACF on z/OS. Every finding hands your engineer the MQSC fix in execution order. |
| Baselines and recurring problems | Not described in IBM's published materials; check with IBM. | Per-object baselines learned from history. Recurring findings are fingerprinted and flagged as chronic, with a permanent-fix recommendation instead of the same temporary fix again. |
| Pricing model | Part of your IBM MQ Advanced or Cloud Pak for Integration entitlement.3 | Priced per estate: queue managers in scope, platforms and deployment model, quoted in writing. The pilot has a fixed fee, credited in full toward your first year if you continue. Model usage is billed to your own provider account. How pricing works |
| Best for | Teams on MQ Advanced who want a conversational assistant to investigate and explain what is happening, on demand. | Teams who want to be warned before an outage, on any MQ edition, including mixed distributed and z/OS estates. |
Not either-or
Using both
Already on MQ Advanced? Use IBM's agent for questions, and let Abendly watch so you don't have to ask at 3am.
- Asking is useful. When you already know something is wrong, a conversational agent can speed up the investigation.
- Watching is better. Most MQ outages announce themselves early: a consumer stops committing, a queue starts to climb. The page is the failure. Everything before it was the warning.
- Neither replaces your engineers. Both advise; a person decides and runs the fix.
Questions
Asked plainly.
Is Abendly a replacement for IBM MQ AI Agents?
No. They do different jobs. IBM MQ AI Agents investigate and explain when an administrator asks. Abendly is built to watch continuously and warn before something breaks. Teams already on IBM MQ Advanced can use both.
Do I need IBM MQ Advanced to use Abendly?
No. Abendly reads queue managers with standard inquire and display authority and MQ event queues, so it does not depend on IBM MQ Advanced.
Does either product change my queue managers?
Neither makes changes for you. IBM describes its agents' queue-manager access as read-only, with administrators in control of execution. Abendly holds inquire and display authority only, enforced by OAM on distributed and RACF on z/OS, and hands your engineer the MQSC fix to review and run.
Can I buy the Abendly monitoring agent today?
The monitoring agent is available as a design-partner pilot on one queue manager in your environment. The pilot has a fixed fee, credited in full toward your first year if you continue. The log analyser and the one-time MQ health check are available now. General availability at full estate scale is on the roadmap.
Which AI model does Abendly use?
Yours. The agent uses your own provider and key: Anthropic, Azure OpenAI, Amazon Bedrock, IBM watsonx, or a self-hosted model. Identifiers are replaced with stable labels inside your environment before anything is reasoned over. How that works
Sources
Where the IBM facts come from.
- IBM MQ AI Agents to accelerate issue resolution and improve messaging resilience, IBM announcement.General availability on 24 March 2026; a component of IBM MQ Advanced, including in Cloud Pak for Integration; supported deployments; up to 20 queue managers per agent instance; diagnosis of message build-up, channel failures and dead-lettered messages; Supervisor Agent as the single entry point.
- IBM MQ AI Agents: Messaging for the autonomous era, IBM product blog.A conversational interface for diagnosing and troubleshooting MQ issues.
- IBM MQ Agent, IBM MQ 10.0 documentation (PDF).Entitlement to the IBM MQ Agent is an extension of the entitlement purchased for IBM MQ Advanced or IBM Cloud Pak for Integration.
- IBM MQ Agent: a technical deep dive, IBM Community, 28 April 2026.Natural-language requests from the agent chat extension in the MQ Web Console (9.4.5 onwards); use of large language models; queue-manager access through a read-only MCP server; installation with Helm charts on Red Hat OpenShift.
- IBM MQ 10.0.1: Advancing enterprise messaging with AI, MQTT 5.0 and enhanced security, IBM announcement.Natural-language MQSC command generation.
- IBM MQ Agents: Detect the Signal. Diagnose the Cause., IBM Community, March 2026.MQ administrators remain in control of execution.
Sources checked on 23 October 2026. IBM's products change; if anything here is out of date or unfair, tell us and we'll correct it.
Asking is useful. Watching is better.
Watch Abendly catch a stalled payments consumer with 14 minutes of runway left, in a simulated estate you can click through.