Skip to main content

Product Office Hours #12 - Jamf @ Jamf: AI Governance Paved Road - Nov 5th

  • Oct 30, 2026 8:52 AM UTC
  • 4 replies
  • 154 views

LysetteB
Forum|alt.badge.img+20

Hi Nation,

Product Office Hours #12 - Jamf @ Jamf: AI Governance Paved Roads

Next session: Thursday, 5th Nov -  8 am PST / 10 am CST / 4 pm GMT / 5 pm CET 
Speakers: Emily Kausalik ( ​@emily ), Manager, Client Platform Engineering
Register here: https://jamf.it/ProductOfficeHours

 
🧵 Got a question? Drop it in the comments below ahead of time if you want it on our radar, though you don't have to. We'll be taking questions live on the call! The Q&A transcript will be added to this thread post event.

See you there!

4 replies

BookMac
Forum|alt.badge.img+16
  • Jamf Heroes
  • Nov 2, 2026 9:00 AM UTC

If i get it right Jamf Protects collects the data from macOS devices for AI Governance. How will it works in the future for iOS and Browser?


emily
Forum|alt.badge.img+26
  • Hall of Fame
  • Nov 5, 2026 8:25 PM UTC

If i get it right Jamf Protects collects the data from macOS devices for AI Governance. How will it works in the future for iOS and Browser?

 

This is all work still in progress, but my sense is anything mobile will tap into Jamf Security Cloud/RADAR/Jamf Trust in combination with Jamf Pro deployment/etc.

For browser, stay tuned for more from my new colleagues from KeepAware: https://keepaware.com/blog/keep-aware-has-been-acquired-by-jamf


emily
Forum|alt.badge.img+26
  • Hall of Fame
  • Nov 5, 2026 8:28 PM UTC

On the call today I mentioned coming up with my own “AI Glossary” for endpoint admins, thought I’d share that here in case anyone else finds it helpful (mainly from the lens of Claude for Claude/Anthropic customers):

 

Terms

1P (first-party): Products and tools that Anthropic builds and runs. Examples include Claude.ai and the Claude API.

3P (third-party): Products or services from other companies. In the Claude context, this usually means Claude models hosted on another cloud, such as Amazon Bedrock, Google Cloud Vertex AI, or Microsoft Foundry. The term can also describe outside vendor tools.

Auto mode (Claude Code): A permission setting in which Claude makes permission decisions for you. A classifier checks each action before it runs and blocks risky ones, such as mass file deletion.

Chat: The back-and-forth Q&A tab in Claude Desktop.

Claude Design: An Anthropic Labs product for creating visual outputs, such as mockups, prototypes, slides, and one-pagers, through conversation with Claude. You can import design systems from a GitHub repository, design files, or uploads. Developers can use the /design command in Claude Code, and designers can use /design-sync to pull in a design system.

Claude Security: An Anthropic product that scans code repositories for vulnerabilities. Findings include confidence and severity ratings, and it can suggest patches. A separate Claude Code plugin, run with /claude-security, does a deep scan locally.

Code: The tab in Claude Desktop for software developers. It also runs in the terminal, IDE extensions, and the browser.

Context window: The amount of text a model can consider at one time. This includes your messages, its replies, and any files or tool results. Long chats can fill it.

Cowork: A task mode in Claude Desktop. You give Claude a multi-step task, and it works through the steps. It can read and write files in folders you choose.

Extended thinking: A mode where Claude works through a problem step by step before it answers. It usually gives better results on hard tasks, but it takes longer and uses more tokens.

Harness: The software around a model that lets it act. It manages tools, the steps the model takes, and the instructions. The model does the reasoning. The harness connects that reasoning to real actions. Claude Code is one example.

Inference: Running a trained model to produce an output. You send a prompt, and the model returns a response.

MCP (Model Context Protocol): An open standard that connects Claude to outside tools and data, such as a calendar, a database, or a ticketing system.

Plugin: A bundle that packages skills, commands, and tools so Claude can do a specific kind of work. Skills are instruction sets. Commands are shortcuts you run. Tools connect Claude to outside systems. Organizations often publish plugins for their own workflows.

Prompt caching: A way to reuse the part of a prompt that does not change, such as a long system prompt or a large document. This lowers cost and speeds up repeat calls.

Skill: A folder of instructions and files that teaches Claude how to do a specific type of task well. Claude loads a skill only when the task needs it.

Token: The unit of text a model reads and writes. A token is roughly a word or part of a word. Usage limits are counted in tokens.

Models

Claude Haiku 5.5: It is the fastest and lowest-cost model in the 5.5 family. Use it for high-volume work, live customer support, and speed-sensitive tasks.

Claude Sonnet 5.5: It is the lower-cost, faster companion to Opus 5.5. Use it for well-scoped everyday work, such as routine coding, bug fixes, document drafting, and high-volume agent tasks.

Claude Opus 5.5: It is the higher-capability model for complex work that needs careful judgment. Use it when a task is ambiguous, long-running, or costly to get wrong.

Claude Fable 5.1 and Claude Mythos 5.1: These are the current Mythos-tier models, which sit above Opus. They share the same underlying model. Fable 5.1 is for general use and includes additional safety measures for biology, cybersecurity, and LLM R&D. Mythos 5.1 has fewer restrictions and is aimed at vetted cybersecurity and life sciences users.

Claude Mythos Preview: It is the first Mythos-class model. Anthropic does not offer it to the public. A small number of trusted organizations use it through Project Glasswing (https://www.anthropic.com/glasswing).

Managed settings via MDM (macOS)

Claude Code managed preferences: /Library/Managed Preferences/com.anthropic.claudecode

Claude Desktop managed preferences: /Library/Managed Preferences/com.anthropic.claudefordesktop

Config file paths (macOS)

Path notes: A path that starts with ~ is in the user home folder, such as /Users/enduser.name. A path that starts with a dot folder, such as .claude/, is relative to the project folder claude is working from (working directory. A path that starts with / is an absolute system path, and it does not depend on your user or project.

Claude Desktop config (home folder): ~/Library/Application Support/Claude/claude_desktop_config.json

  • This file holds MCP server setup.

Claude Desktop developer settings (home folder): ~/Library/Application Support/Claude/developer_settings.json

Claude Desktop logs (home folder): ~/Library/Logs/Claude/

  • Check the mcp*.log files when a connector fails.

Claude Code user settings (home folder): ~/.claude/settings.json.

  • These settings apply to all projects.

Claude Code user state and MCP servers (home folder): ~/.claude.json

  • This file sits directly in your home folder, not inside ~/.claude/.

Claude Code project settings (working directory): .claude/settings.json

  • Inside each project folder. Share this file through source control.

Claude Code local project settings (working directory): .claude/settings.local.json

  • inside each project folder. This file is for contains unique/end user overrides and is usually not shared.

Claude Code memory file, project level (working directory): CLAUDE.md

  • In the project folder. Claude reads it at the start of a session.

Claude Code memory file, user level (home folder): ~/.claude/CLAUDE.md

  • Claude reads it in every project.

Claude Code login: Stored in the macOS Keychain, under the login keychain entry named Claude Code.

Claude Code config folder override: The CLAUDE_CONFIG_DIR environment variable moves the ~/.claude folder to another location.

Resources

Claude Desktop enterprise configuration and MDM keys: https://support.claude.com/en/articles/12622667-enterprise-configuration-for-claude-desktop

Claude Code settings scopes and precedence (managed, command line, local, project, user): https://code.claude.com/docs/en/settings

Permission modes, including auto mode: https://code.claude.com/docs/en/permission-modes

Auto mode announcement: https://claude.com/blog/claude-code-auto-mode

Claude Security plugin for Claude Code: https://code.claude.com/docs/en/claude-security

Cowork: https://support.claude.com/en/articles/12138966

Claude Security product help: https://support.claude.com/en/articles/14661296-claude-security

Project Glasswing: https://www.anthropic.com/glasswing

General developer docs: https://docs.claude.com


BookMac
Forum|alt.badge.img+16
  • Jamf Heroes
  • Nov 6, 2026 6:51 AM UTC

​@emily Really helpful, thanks!