- Published on
IBM Brings Bob to Self-Hosted and Air-Gapped Environments: Agentic Software Development Without Moving Your Code
- Authors
- Name
- DeepHorus
IBM has announced that a self-hosted deployment option for IBM Bob, its agentic software development platform, is now generally available to enterprise customers. The move allows organizations to run Bob on premises, in private or sovereign clouds, and in fully air-gapped networks where public internet access is prohibited. Bob covers the entire software development lifecycle, from understanding code and planning work to executing changes and validating results. With this release, IBM is targeting regulated industries and organizations with strict data residency requirements that have previously been locked out of cloud-based AI development tools.
The self-hosted option ships with the full core Bob experience, including the integrated development environment (IDE), BobShell, parallel tool calling, the agent harness, and support for skills and modes. Customers can also deploy self-hosted, air-gapped, and hybrid model configurations, along with bring-your-own-license (BYOL) arrangements for eligible models on IBM's supported list. Optional Premium Packages extend Bob to Java Modernization, IBM i, and IBM Z workloads, targeting legacy estate modernization. IBM has not published pricing for the offering and instead directs prospective buyers to request a demo.
A defining characteristic of Bob self-hosted is that it does not bundle a model with the platform. Customers must source, license, and host a supported model themselves, choosing from IBM's list grouped by deployment method. This architectural decision determines where inference runs and, consequently, where code, development context, and build artifacts are processed. With a self-hosted model such as Nemotron or Poolside Laguna, all data can remain within the customer's own environment, offering full isolation for sensitive workloads.
IBM illustrates the flexibility of this approach with a banking scenario. A bank could run on-premises inference for core banking applications while routing less restricted workloads to an approved external model such as Claude, Gemini, or GPT through a hybrid configuration. Developers experience a consistent core Bob interface in both cases, while platform and security teams retain control over the underlying architecture. IBM has also signaled plans to expand its model portfolio and introduce multi-model routing, though these remain on the roadmap rather than shipped features.
The announcement differentiates IBM from rivals such as GitLab and GitHub, which embed agents within their own DevOps platforms, and Mistral, which pairs its agent with proprietary open-weight models. IBM's focus on modernizing long-lived estates, particularly Java, IBM i, and mainframe code, inside environments that cannot touch the public internet positions it distinctly in the market. The offering is available today for enterprise customers with supported infrastructure environments.
Looking ahead, the self-hosted release signals IBM's intent to make agentic AI development accessible to security-conscious enterprises. As organizations increasingly demand control over where their code is processed, Bob self-hosted offers a path to adopt AI-assisted development without compromising compliance or sovereignty.
