Canonical’s Workshop tackles developer experience and agentic AI sandboxing, standardising environments via one command.
These environments are configured once and can then be reproduced across entirely different machines. This ensures consistent workflows spanning from individual development hardware directly through to deployment pipelines; requiring less time spent managing complex dependencies.
Platform engineering leads continuously search for methods to reduce the cognitive load on their teams. Workshop addresses this by allowing developers to define their environments within simple YAML documents. Because these configurations exist as plain text, they can be version controlled and shared easily among project contributors. Creating, upgrading, or winding down an environment can be accomplished with just a few keystrokes.
Jon Seager, VP of Engineering at Canonical, said: “Developers operating at the cutting edge want to focus on what they’re building, not on dependencies or workstation configuration.
“Workshop enables developers to achieve that elegantly with a single YAML file that defines their environment, and pulls the exact dependencies and components they need. Workshops also serve to standardise and sandbox agentic tooling consistently across teams.”
Streamlining hardware acceleration and SDK integration
Modern engineering workloads increasingly depend on machine learning models and hardware acceleration. Using Workshop, individuals can pull in required SDKs – such as Ollama, OpenCode, NVIDIA CUDA, and AMD ROCm – by simply including them within their Workshop configuration files. This declarative approach replaces fragile shell scripts and lengthy manual installation wikis that often plague developer onboarding.
If a pre-designed SDK already exists to access a specific piece of hardware, engineers can choose to incorporate it into their YAML, allowing them to optimise the resulting toolchain for their specific machine. If an SDK does not exist, developers have the flexibility to create a custom one tailored to their specific needs.
This treats hardware passthrough as a first-class configuration object. Data science teams and platform engineers can share identical base configurations without forcing the platform team to maintain heavy container images packed with every conceivable library.
Integrating these accelerators often causes friction due to driver version mismatches between the host operating system and the container payload. Workshop attempts to address this by defining the explicit dependencies required for the container to interact with host hardware securely.
By forcing a declarative configuration style, engineering departments can reduce onboarding time for new hires. The abstract nature of the interface system also means that as underlying host operating systems undergo updates, the configuration file remains static.
Navigating agentic AI sandboxing
Teams building agentic AI require resilient sandboxing to ensure they can develop at speed, without risking harm to the underlying host system. Workshop addresses this security vector by running development environments strictly within unprivileged system containers. This architectural choice serves to minimise the attack surface available to workloads operating within them.
Repeatability within this framework applies directly to security postures. In Workshop, SDKs are limited to a uniform means of requesting access to host resources, such as requiring access to the desktop to display a GUI app, or requesting access to the SSH agent located on the host machine. Security enforcement does not depend on knowing the individual protocols of a container, meaning strict access controls can be enforced for AI agents.
Dmitry Lyfar, Engineering Manager at Canonical, explained: “Ease of use for developers shouldn’t mean ease of access for AI agents. There’s naturally a tension between these two ‘user’ groups, but Workshop resolves it through strict enforcement of access controls.
“Resource allocation remains simple and consistent across all environments to minimise human error, while non-privileged defaults effectively constrain workload capabilities.”
Integration constraints for platform teams
Ensure you are running LXD 6.8 or newer before installing Workshop. Installation mandates the use of snap packages, requiring engineers to install LXD via the stable channel and Workshop via the classic confinement channel.
For organisations invested in alternative setups like devcontainers or Nix flakes, migrating to an LXD-based workflow requires careful evaluation. Continuous integration and deployment pipelines must be updated to support this new format.
CI environments are often ephemeral and tightly constrained. Running nested system containers introduces specific networking and storage driver complexities. Platform architects must evaluate whether the benefit of a unified YAML definition across local and CI environments outweighs the initial engineering effort required to provision LXD-compatible runner infrastructure.
Integrating Canonical Workshop into existing enterprise architectures requires addressing backwards compatibility. Legacy applications often depend on highly specific host configurations that are difficult to containerise without extensive refactoring.
While Workshop provides a path forward for greenfield projects and modern microservices, platform teams will need to build migration strategies for older monolithic applications. The use of unprivileged containers provides a strong boundary, but legacy applications attempting to make privileged system calls will fail securely by default. Engineering teams will need to profile these applications and map out exactly which system calls are necessary, subsequently defining those explicit permissions within the Workshop YAML.
Managing dependency caching becomes a focal point when hundreds of developers begin pulling SDKs simultaneously. Because developers can pull heavy frameworks directly via configuration files, enterprise networks could experience congestion. Platform engineering teams will need to implement local caching proxies or internal mirrors for these SDKs to ensure that the single-command environment provisioning remains fast and reliable.
Rethinking host resource allocation
Managing local hardware passthrough in containerised environments typically involves brittle volume mounts and complex network bridge configurations. Workshop dispenses with complex mapping scripts and filesystem paths entirely. Instead, the tool offers a standardised way to access mounts, devices, and network services directly from the contained environment.
This abstraction is managed through an interface system that is inspired by snapd, which simplifies host resource allocation. By abstracting the underlying Linux kernel namespaces and cgroups into a more digestible format, developers can focus on application logic rather than debugging container networking protocols.
Zooming out to the broader cloud-native ecosystem reveals a deliberate architectural parallel with Kubernetes. Platform operators already rely on declarative YAML to dictate the exact state of production clusters, and Workshop applies that same operating logic directly to the local machine.
The mental friction of moving between development and operations tasks drops off when both domains share a configuration dialect. This shared approach bridges the traditional divide between application engineers and infrastructure teams.
Offloading the heavy lifting of workstation maintenance onto Canonical allows internal platform engineers to stop building bespoke provisioning tools and focus entirely on system reliability.
See also: NVIDIA CUDA 13.3 bridges the Python and C++ divide for AI teams
Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is part of TechEx and is co-located with other leading technology events including the Cyber Security & Cloud Expo. Click here for more information.
Developer is powered by TechForge Media. Explore other upcoming enterprise technology events and webinars here.