IBM Brings Bob to Self-Hosted and Air-Gapped
IBM has launched self-hosted Bob for enterprise customers, enabling agentic software development without moving code outside secure environments.

Stock photo for illustration only, not from the actual event
- IBM's self-hosted Bob is now generally available for enterprise customers.
- Enables agentic software development while keeping code and context on-premises.
- Customers must source, license, and host a supported model themselves.
IBM has made a major move by bringing its agentic software development tool, Bob, to self-hosted and air-gapped environments. This capability allows organizations to leverage advanced AI assistance without needing to transmit their source code outside secure perimeters, with the self-hosted option now generally available to enterprise customers.
This approach addresses strict security constraints by requiring customers to source, license, and host a supported model themselves. IBM has not published formal pricing details for this deployment method and currently directs prospective buyers to submit demo requests.

Stock photo for illustration only, not from the actual event
When deployed in a self-hosted setup, Bob does not come bundled with a default model. Instead, customers must bring their own from IBM's approved list, which is organized by specific deployment methods:
- Processing code and development context entirely inside the customer environment.
- Integrating seamlessly with existing highly secure enterprise infrastructure.
- Ensuring that sensitive source codes and build artifacts never leave the local network.
Bringing Bob to self-hosted and air-gapped setups marks a strategic push by IBM to capture large enterprise clients, such as banks and government agencies, bound by strict data residency and compliance regulations. Allowing developers to utilize AI coding assistants in closed networks removes a major barrier to adopting modern developer tooling.
IBM illustrates this capability with a practical example: a bank could run on-premises inference for core banking workloads while routing less restrictive tasks to an approved external model. Developers maintain a consistent core Bob experience across both scenarios, while platform and security teams retain strict control over the underlying architecture.
Looking ahead, IBM plans to expand its supported model portfolio and introduce multi-model routing capabilities. However, these upcoming enhancements remain part of the product roadmap rather than currently shipped features.
The primary differentiator for IBM lies in its targeted scope. While platforms like GitLab and GitHub integrate agents directly into their proprietary DevOps ecosystems, and Mistral pairs its agents with open-weight models, IBM focuses heavily on modernizing legacy enterprise estates—particularly Java, IBM i, and mainframe code—inside isolated environments completely disconnected from the public internet.
Source: MarkTechPost
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment