Yes. IBM Bob self-hosted is generally available for deployment on a customer-managed Red Hat OpenShift cluster, either on premises or in the customer’s own cloud account. That does not automatically keep source code or inference inside the organization’s network: this depends on whether Bob uses a locally hosted model or sends requests to an approved external model service.
What “self-hosted” means for IBM Bob
In this deployment, Bob’s backend runs on infrastructure the customer operates rather than IBM’s SaaS infrastructure. IBM describes the self-hosted offering as providing Bob’s core IDE experience, BobShell, parallel tool calling, agent harness, skills, and modes in supported configurations. The deployment uses the Bob Kubernetes Operator and Helm charts, with operator and instance namespaces on OpenShift. IBM says Bob can share a cluster with other workloads if it has sufficient resources; a dedicated cluster is not required. IBM’s self-hosted overview and its October 2026 release post describe the architecture.
| # | Preview | Product | Price | |
|---|---|---|---|---|
| 1 |
|
IBM's Workstation Cics: Servers, Workstations, and Lan : Pc-DOS and Os/2 : Distributed Processing... | $36.06 | Buy on Amazon |
Does source code stay on your network?
It depends on the model connection. Bob’s backend location and the location of model inference are separate decisions: self-hosting the backend does not by itself make model use local.
| Configuration | Where inference happens | What happens to request context |
|---|---|---|
| Hybrid with an external model service | The selected external service | Model requests, including code context, go to the organization’s chosen service under its agreements. Bob’s backend, identity, audit logs, and usage metering remain on the customer’s cluster, according to IBM’s release post. |
| Fully self-hosted | On the organization’s infrastructure, using its GPUs | Inference can run locally. IBM names NVIDIA Nemotron 3 Ultra and Poolside Laguna S 2.1 for this route and describes a small guardrail model screening inputs and outputs. |
IBM’s release post is the source for the named models and the distinction between these deployment paths. Model selection, workload configuration, and any applicable agreements determine the actual data flow; organizations should verify those details against their own deployment and model-service terms before using sensitive code.
#1 Best Overall
Can IBM Bob run in an air-gapped environment?
IBM says the self-hosted path supports disconnected installation. Its October 2026 release post describes using bobctl to mirror images into a private registry, including transferring them from a connected system across an air gap. IBM also says a single-node OpenShift installation can support a proof of concept. That is not a production sizing recommendation: production requirements depend on the deployment and workload.
What does deployment require?
OpenShift and capacity
The self-hosted backend requires a customer-managed Red Hat OpenShift Container Platform cluster. It may share a cluster with other workloads if resources are adequate. For fully self-hosted model inference, the organization also supplies GPU infrastructure. IBM says sizing depends on factors including quantization, context length, and the number of concurrent developers; the reviewed public materials do not establish a universal minimum production configuration. See the deployment overview and release post for IBM’s deployment guidance.
Installation and identity
IBM describes installation through the Bob Kubernetes Operator and Helm charts. Corporate identity can be configured through LDAP or Active Directory; smaller installations may manage users directly in the included identity service. For disconnected deployment, IBM’s release post describes mirroring images with bobctl into a private registry.
What does the customer operate?
Self-hosting moves operational work to the customer. The organization is responsible for the backend’s lifecycle operations and for configuring networking, storage, and identity. IBM’s overview also says security event logging and monitoring are handled at the OpenShift platform level and are not provided by Bob. Teams evaluating the deployment should account for these responsibilities alongside their OpenShift administration and, where applicable, GPU and model operations.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →How does self-hosted compare with IBM Bob SaaS?
| Consideration | IBM Bob SaaS | IBM Bob self-hosted |
|---|---|---|
| Infrastructure and upgrades | IBM manages the infrastructure and upgrades. | The customer operates the backend and manages its lifecycle on its OpenShift cluster. |
| Model location and request context | Depends on the model connection and configuration. | Can use customer-hosted inference, or a hybrid external service that receives model requests and code context. |
| Disconnected installation | Not described as a disconnected deployment option in the reviewed IBM materials. | IBM says disconnected installation is supported, with image mirroring into a private registry. |
| Identity and security logging | IBM manages SaaS infrastructure; consult IBM’s service terms for the specific operating boundary. | The customer configures identity and platform security logging and monitoring. |
| Platform requirement | No customer-managed OpenShift requirement is identified for SaaS in the reviewed sources. | Requires a customer-managed Red Hat OpenShift Container Platform cluster. |
IBM’s overview of self-hosted Bob sets out the operating split. Model location is not a simple SaaS-versus-self-hosted distinction: a self-hosted Bob backend can still use an external model service.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Availability, access, and licensing
IBM’s October 1, 2026 release post says self-hosted general availability took effect September 24, 2026; IBM published a separate availability announcement on September 30. The route to access is sales-led: contact IBM or a Business Partner, or use Contact Sales, after which an account team arranges organization entitlement. IBM’s reviewed public sources do not publish self-hosted pricing or exact enterprise entitlement terms. The release post has the access details, and IBM’s announcement covers availability and package options.
IBM identifies optional Premium Packages for Java Modernization, IBM i, and IBM Z. Availability is subject to applicable licensing and deployment conditions; organizations should confirm eligibility and terms with IBM or a Business Partner.
Why older IBM pages may say on-premises was still planned
IBM’s April 28, 2026 SaaS general-availability release described on-premises deployment as a future target. That statement reflected the position at the time and was superseded by IBM’s September and October 2026 self-hosted announcements. The April release is useful historical context, not the current availability status.
Do these 3 things before closing this tab:
1Fix the driver behind crashes, sound loss and screen glitches2Repair Windows errors before they cause bigger problems3Scan for outdated or missing drivers - takes under a minuteQuick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
