Comprehensive guides and documentation to help you start and utilize our platform.
The dell-ai agent skill teaches compatible AI coding agents how to work with Dell Enterprise Hub. With the skill installed, you can describe a task in natural language and your agent can select the appropriate dell-ai CLI or Python SDK workflow, check the required inputs, and help you run it.
The skill is an instruction layer for your agent. It does not replace the dell-ai package, your Hugging Face authentication, or the local deployment tools used by a generated command.
The agent uses the local dell-ai package to interact with Dell Enterprise Hub. Create a virtual environment and install it:
uv venv .venv
source .venv/bin/activate
uv pip install dell-aiThen authenticate with your Hugging Face account:
dell-ai login
dell-ai whoamiYour token must have access to the models you want to use. For gated models, you must also accept the model author's access conditions on the Hugging Face Hub.
Use the dell-ai skills commands to install and inspect the skill. Choose the destination for your agent and whether the skill should be available only in the current project or globally.
Install the skill for Devin in the current project:
dell-ai skills add dell-ai --dest .devin/skillsOr install it for Cursor globally:
dell-ai skills add dell-ai --cursor --globalView the installed skill:
dell-ai skills show dell-aiStart a new agent session after installation so the agent discovers the skill. The canonical skill definition is available in the dell-ai GitHub repository.
Ask the agent to search the Dell Enterprise Hub catalog, compare model metadata, find platforms that support a model, or inspect applications and their configurable parameters.
Find the multimodal models available for
xe9680-nvidia-h200, then compare their size and license.
Show me which Dell platforms support
google/gemma-3-27b-it.
The agent can generate validated Docker or Kubernetes deployment snippets for a compatible model and platform. It can use a GPU count you provide or request a goodput-optimized configuration for balanced, long-context, high-concurrency, or performance workloads.
Generate a Docker deployment command for
google/gemma-3-27b-itonxe9680-nvidia-h200using 8 GPUs. Do not run it.
Create a Kubernetes snippet optimized for a high-concurrency workload and explain the selected configuration.
When you ask it to execute a deployment, the agent can use Docker or Kubernetes for models and Helm for catalog applications. It can then inspect endpoint health, list tracked deployments, and tear down a specific deployment.
Deploy
google/gemma-3-27b-itwith Docker, then report the endpoint and deployment ID.
Check the status of my Dell AI deployments and remove the stopped one.
Running workloads requires the relevant local engine (docker, kubectl, or helm) to be installed and configured. Review the proposed action before authorizing commands that deploy or remove resources.
On Linux, the agent can describe the local hardware and compare it with Dell AI validated configurations. It can also inspect checkpoints and active Docker or Kubernetes workloads through dell-ai status.
Check whether this system matches a Dell AI validated configuration and summarize any incompatibilities.
The agent can set, inspect, and remove dell-ai environment values at project or global scope, including checkpoint locations reported by status checks.
Set
DELL_AI_CHECKPOINTfor this project and verify that dell-ai can see it.
The skill guides the agent to validate model access and model-platform compatibility before generating a deployment. It also teaches the agent the difference between generating a snippet and executing it, and how to handle deployment IDs, container image tags, local model weights, goodput scenarios, and common API errors.
The Dell Enterprise Hub API remains the source of truth for current models, platforms, applications, container tags, and optimized configurations. Results can therefore vary as the catalog changes.