How to use the Arize AI Provider Integration agent skill

What is an agent skill and what does this one do?

An agent skill is a folder containing a SKILL.md instruction file and optional reference files. When installed, the agent reads these instructions when a request matches the skill's description, and then follows them, including running commands on your behalf.

This skill covers AI integrations: stored LLM provider credentials registered in Arize. Evaluators use these integrations to call a judge model, and other Arize features use them when they need to invoke an LLM. With the skill, the agent can run the ax command-line tool to list integrations, look one up, create a new one, update credentials or delete it.

• List integrations in a space, optionally filtered by name
• Get the full configuration of one integration by name or ID
• Create integrations for OpenAI, Anthropic, Azure OpenAI, AWS Bedrock, Vertex AI, Gemini, NVIDIA NIM or a custom OpenAI-compatible endpoint
• Capture the integration ID needed for evaluator creation
• Troubleshoot common failures such as a missing profile or API key

What do you need before installing?

The skill does not install the Arize tooling for you. You need the Arize ax CLI installed and available on your PATH, and a configured Arize profile that contains an API key. The API key is created on the admin page at app.arize.com. You also need the LLM provider API key, or the cloud role or project details, for whichever provider you want to connect.

Every command needs a configured Arize profile, so you need an Arize account before this skill is useful.

• Arize ax CLI installed and on PATH
• A configured Arize profile with an API key (created at app.arize.com, admin page)
• LLM provider API key, or cloud role or project details for the provider you are connecting

How do you install the skill?

Download the skill ZIP from the page where you found this guide and unzip it. For a personal install, place the resulting folder in ~/.claude/skills/. For a single project, place it in that project's .claude/skills/ folder.

In the Claude apps you can instead upload the skill under Settings → Capabilities → Skills.

• Personal install: unzip into ~/.claude/skills/
• Project install: unzip into the project's .claude/skills/ folder
• Claude apps: Settings → Capabilities → Skills, then upload

How do you trigger the skill with a natural request?

You do not need to name the skill. The agent uses it when your request mentions AI integration, LLM provider credentials, creating, listing, updating or deleting an integration, or connecting an LLM provider to Arize. Name the space if you know it, since most commands need one.

• "List the AI integrations in my space called production."
• "Create an Anthropic integration in Arize called Eval Judge using my key from the environment."
• "Find the OpenAI integration and show me its configuration."
• "Connect AWS Bedrock to Arize using this role ARN."
• "Delete the old Gemini integration."

What does a typical command look like?

The skill tells the agent to list integrations before creating one, in case a suitable one already exists. A basic listing command looks like this. Pagination is available with --limit and --cursor, and -o json gives machine-readable output.

Creating an integration uses the provider name in the --provider flag. For AWS Bedrock and Vertex AI, the skill passes details through --provider-metadata instead of an API key.

What safety points should you know about?

The skill runs scripts and uses the network, because the ax CLI sends requests to Arize. Its instructions tell the agent to never read .env files or search the filesystem for credentials, and to use ax profiles for Arize credentials and ax ai-integrations for provider keys.

Do not paste provider API keys into an agent chat. Keys then end up in shell commands and transcripts. Store them in ax profiles or a secrets manager, and reference environment variables instead of raw key values.

• Runs scripts: yes
• Uses network: yes
• Reference keys through environment variables, never as raw values in chat
• Use ax profiles for Arize credentials and a secrets manager for provider keys where possible

What tips help when things go wrong?

If a command fails, the skill directs the agent to the right reference file. A command-not-found or version error points to the setup notes, and a 401 Unauthorized or missing API key error points to the profile check. Ask the agent to run ax profiles show to see the current profile state, and let it fix only what is broken.

If you do not know your space, run ax spaces list. Space flags accept either a name or a base64 ID. Note that ai-integrations create does not accept --space, because AI integrations are account-scoped.

• Space flags accept a space name or a base64 space ID
• ai-integrations create does not take --space; list, get, update and delete do
• Save the integration ID returned on creation for later evaluator setup