Coding agent setup

Use AI coding agents like Cursor, Claude Code, and Copilot to set up and work with Arize AX via skills, MCP servers, or a single URL.

You’re using a coding agent — Cursor, Claude Code, Copilot, Windsurf, or something else — and you want it to know how to work with Arize AX. Here’s how:

• Paste-and-Go — paste a single URL, agent does the rest
• Skills — teach your agent the full Arize workflow: tracing, datasets, experiments, evals, prompt optimization, and more
• MCP Servers — persistent IDE integration for instrumentation help and doc lookups
• Copy Docs — paste any docs page as markdown into your agent for context

---

Paste-and-Go Tracing Setup

The fastest path. Paste this into your coding agent and it handles dependency installation, instrumentation, and configuration:

Paste this into your coding agent:

Follow the instructions from https://arize.com/docs/PROMPT.md and ask me questions as needed.

The agent scans your codebase, detects your language and frameworks, proposes an instrumentation plan, and implements it once you approve. Supports Python, TypeScript/JavaScript, and Java across 30+ integrations.

For more control, use the arize-instrumentation skill instead.

---

Skills

Arize Skills encode the workflows we’ve refined building the Arize AX platform and helping teams debug LLM apps in production. They handle the ax CLI flags, data shape quirks, and multi-step recipes so your agent doesn’t have to guess. Your agent gets the full Arize workflow — add tracing, export traces, create datasets, run experiments, set up evaluators, manage annotations, optimize prompts — all through natural language.

Works with Cursor, Claude Code, Codex, Windsurf, and 40+ other agents.

Already have traces? Give your agent this prompt to install skills and start debugging:

Install the Arize skills plugin from https://github.com/Arize-ai/arize-skills, then use the arize-trace skill to export and analyze recent traces from my project. Summarize any errors or latency issues you find.

Prerequisites

1. AX CLI:

   uv tool install arize-ax-cli   # or: pipx install arize-ax-cli
   

2. Authentication — set up your API key using one of these methods:

   ax profile (recommended)

   # Interactive wizard
   ax profiles create
   
   # Or pass the key directly
   ax profiles create --api-key YOUR_API_KEY
   

   .env file

   cp .env.example .env
   # Edit .env with your credentials:
   # ARIZE_API_KEY=your-api-key
   # ARIZE_SPACE=my-workspace       # space name or base64 ID from ax spaces list
   

   Skills automatically load this file during their prerequisite check.

   Environment variables

   export ARIZE_API_KEY="your-api-key"
   export ARIZE_SPACE="my-workspace"   # space name or base64 ID from ax spaces list
   

3. Space — ARIZE_SPACE accepts a space name (e.g. my-workspace) or base64 ID. Find yours with ax spaces list or from the Arize URL (/spaces/{SPACE_ID}/...).

4. Verify:

   ax --version && ax profiles show
   

See AX CLI overview for full details.

Install Skills

Option 1: npx (recommended)

# Interactive — choose skills, agent, and scope
npx skills add Arize-ai/arize-skills

# Non-interactive — install everything
npx skills add Arize-ai/arize-skills --skill '*' --yes

Option 2: ax CLI

If you already have ax installed (v0.9.0+):

ax skills install

Option 3: Claude Code plugin

/plugin marketplace add Arize-ai/arize-skills
/plugin install arize-skills@arize-skills

Option 4: git clone

macOS / Linux

git clone https://github.com/Arize-ai/arize-skills.git
cd arize-skills
./install.sh --project ~/my-project

Windows (PowerShell)

git clone https://github.com/Arize-ai/arize-skills.git
cd arize-skills
.\install.ps1 -Project ~\my-project

The installer auto-detects your agents and optionally installs the ax CLI. Use --global / -Global to install to ~/.<agent>/skills/. Run --list to see available skills, --uninstall to remove them.

Updating skills: npx skills update or cd arize-skills && git pull.

Available Skills

Skill	What it does
arize-instrumentation	Add Arize AX tracing to an app — two-phase flow: analyze codebase, then implement instrumentation
arize-trace	Export traces and spans by trace ID, span ID, or session ID; debug LLM application issues
arize-dataset	Create, manage, and download datasets and examples
arize-experiment	Run and analyze experiments against datasets
arize-evaluator	Create LLM-as-judge evaluators, run evaluation tasks, and set up continuous monitoring
arize-ai-provider-integration	Create and manage LLM provider credentials (OpenAI, Anthropic, Azure, Bedrock, Vertex, and more)
arize-annotation	Create and manage annotation configs (categorical, continuous, freeform); bulk-annotate project spans
arize-prompt-optimization	Optimize prompts using trace data, experiments, and meta-prompting
arize-prompts	Manage Prompt Hub templates and versions with ax prompts (create, label, promote)
arize-link	Generate deep links to traces, spans, and sessions in the Arize AX UI
arize-admin	Manage users, orgs, spaces, roles, role bindings, and API keys — enterprise access control and team onboarding
arize-compliance-audit	Audit an AI app for regulatory compliance (EU AI Act, NIST AI RMF, GDPR, HIPAA) — produces a tailored remediation checklist

---

MCP Servers

Two MCP servers you can plug into your IDE for persistent access to instrumentation guidance and documentation.

Server	Best for	What it does
Tracing Assistant (PyPI)	Instrumentation	Framework-specific examples, code analysis, direct Arize support
Docs MCP Server	Reference	Search all Arize AX docs, examples, and API references from your IDE

Install uv

The Tracing Assistant requires uv (a fast Python package manager). The Docs Server needs no installation.

macOS

pip install uv  # or: brew install uv

Linux

curl -LsSf https://astral.sh/uv/install.sh | sh

Windows

powershell -c "irm https://astral.sh/uv/install.ps1 | iex"

Add to Your IDE

Cursor

Navigate to Settings → MCP → Add new global MCP server, then add the JSON config:

Tracing Assistant

"arize-tracing-assistant": {
  "command": "uvx",
  "args": ["arize-tracing-assistant@latest"]
}

Docs MCP Server

"arize-ax-docs": {
  "url": "https://arize.com/docs/mcp"
}

Claude Code

Tracing Assistant

claude mcp add arize-tracing-assistant uvx arize-tracing-assistant@latest

Docs MCP Server

claude mcp add arize-ax-docs --transport http https://arize.com/docs/mcp

Verify with claude mcp list.

Gemini CLI

Tracing Assistant

gemini extensions install https://github.com/Arize-ai/arize-tracing-assistant

Docs MCP Server

gemini mcp add arize-ax-docs https://arize.com/docs/mcp

Antigravity (Google)

Open the MCP Store panel (from the ... dropdown in the editor’s side panel), search “Arize”, and click Install. Alternatively, add the Tracing Assistant JSON config manually via Manage MCP Servers → View raw config.

Claude Desktop / Manual Config

Add to your mcpServers JSON config (Settings → Developer → Edit Config in Claude Desktop):

Tracing Assistant

"arize-tracing-assistant": {
  "command": "uvx",
  "args": ["arize-tracing-assistant@latest"]
}

Docs MCP Server

"arize-ax-docs": {
  "url": "https://arize.com/docs/mcp"
}

Troubleshooting: Run uv cache clean to reset cached versions, or start the server manually with uvx arize-tracing-assistant@latest to see errors. Debug with npx @modelcontextprotocol/inspector uvx arize-tracing-assistant@latest.

---

Copy Docs to Your Agent

Every page in these docs has a copy as markdown button — paste it directly into ChatGPT, Claude, or any coding agent for accurate, up-to-date context.