Discover and deploy AI models with the Arm AI Portal MCP server
Introduction
Connect your AI harness to the Arm AI Portal MCP server
Find models and deployment paths with natural-language prompts
Next Steps

  Log an issue

  Fork and edit

  Discuss on Discord

About this Learning Path

Skill level:  Introductory
Reading time:  20 min
Last updated:  07 Sep 2026
Skill level:
  Introductory
Reading time:
  20 min
Last updated:
  07 Sep 2026
Author:Pinru Wang
Arm IP:Neoverse
Tags: ML Linux macOS Windows MCP Arm AI Portal Arm Topo
Author:
  • Pinru Wang
Arm IP:
Neoverse
Tags:
ML Linux macOS Windows MCP Arm AI Portal Arm Topo

Who is this for?

This Learning Path is for developers who want to use the Arm AI Portal Model Context Protocol (MCP) server to discover models and plan deployments for Arm-based cloud or edge targets.

What will you learn?

Upon completion of this Learning Path, you will be able to:

  • Connect the Arm AI Portal Model Context Protocol (MCP) server to an MCP-compatible AI harness
  • Search and compare models by task, runtime, performance, memory usage, and Arm target, and find relevant documentation
  • Deploy and validate a selected model on an Arm-based edge target
  • Identify the supported workflow for cloud deployment and the current limits of mobile deployment

Prerequisites

Before starting, you will need the following:

  • An MCP-compatible AI client, such as Codex, Claude Code, or GitHub Copilot
  • Basic familiarity with AI model tasks, runtimes, and deployment targets
  • For edge deployment, a Docker-capable arm64 Linux device reachable over SSH

Summary

AI-assisted

This summary was drafted with an approved AI-assisted workflow and reviewed by Arm contributors before publication. Human technical review remains part of the process so the final page reflects engineering rigor, accuracy, and Arm editorial standards.

Close
?
You’ll connect an MCP-compatible AI harness to the Arm AI Portal MCP server using the server’s URL. After setting up the MCP server, you’ll search the model catalog with natural-language prompts and compare candidates by task, runtime, latency, performance, memory usage, and Arm target. You’ll also find related documentation and explore deployment paths for cloud and Arm-based edge targets. Finally, you’ll learn best practices for deploying with the MCP server.

Frequently asked questions

AI-assisted

These FAQs were drafted with an approved AI-assisted workflow and reviewed by Arm contributors before publication. Human technical review remains part of the process so the final page reflects engineering rigor, accuracy, and Arm editorial standards.

Close
?
Do I need to install the Arm AI Portal MCP server locally?
No. The server uses Streamable HTTP. Add it to your AI harness by specifying the MCP server’s URL.
How do I know the MCP server is connected to my AI harness?
List the configured MCP servers and confirm that arm-ai appears as connected. Then, ask your harness to use the Arm AI Portal MCP server and call find_model. If it uses web search instead, explicitly ask it to use the arm-ai MCP tools.
What kind of query helps me narrow to the right model?
Start with your goal, then add constraints such as model task, Arm target, runtime, latency, memory, and performance. If you’re unsure of the task, begin with a broad question and refine based on the returned options.
How do I use the MCP server to plan a deployment?
Prompt your harness the model and target that you selected and how you can access the target. Ask it to separate prerequisites, target preparation, deployment, validation, and cleanup. Review the plan and approve it before the harness runs commands or changes the target.
How do I validate a deployment guided by the MCP server?
Ask your harness to check the deployment status and run a representative inference test. Use input and success criteria that match the model task. A running process or successful network response doesn’t prove that the model produces a correct result.
Next