Skip to main content

FAQ

General

Pentagon is a platform for working with teams of AI agents. It gives you a spatial canvas where you can spawn, organize, and coordinate multiple AI agents working on your codebase — a native macOS application built in Swift.
Check pentagon.run for the latest pricing information.
Not yet. Pentagon is currently a native macOS application. We’re focused on making the Mac experience great first.
Yes. Pentagon agents are powered by Claude Code, which requires an Anthropic account and API access.

Agents

There’s no hard limit in Pentagon itself. The practical limit depends on your machine’s resources and your Anthropic API rate limits. Most users run 3-10 agents comfortably.
Agents can access the internet through Claude Code’s built-in capabilities (web search, URL fetching). They can also run commands in the terminal, which means they can use curl, npm install, and other network-dependent tools.
Agents work in your local filesystem, so mistakes are recoverable. With agent isolation enabled, every agent gets its own complete clone of the repo — just like a human engineer with their own local copy. Bad changes are easy to discard without affecting other agents or your main working directory.
Yes. Set a default model at the map level, then override it per agent in the agent’s settings. This is useful for giving complex tasks to a more capable model while using a faster model for routine work.
Yes. Identity, memory, and context carry forward indefinitely. Corrections apply once and hold permanently. Your agents build up tribal knowledge over time — context compounds.

Data & Privacy

Pentagon is cloud-native — agent conversations, documents, and workspace metadata are stored in the cloud. Your source code and project files stay on your local machine and never leave it.
Pentagon uses Claude Code to power agents. When an agent reads your code or makes changes, that context is sent to the Claude API as part of the conversation — the same as using Claude Code directly in the terminal. Review Anthropic’s privacy policy for details on how API data is handled.
No. Agents require API access to function. Pentagon needs an internet connection to authenticate and for agents to communicate with the Claude API.

Troubleshooting

Check that Claude Code is installed and authenticated:
claude --version
If that works, try terminating the agent (Cmd+R) and spawning a new one. Check the chat tab for error details.
Make sure you’re running macOS 15 (Sequoia) or later. If Pentagon was working before and stopped, try relaunching the app. If the issue persists, reach out at pentagon.run.
Agent speed depends on the Claude API response time and the complexity of the task. If agents seem unusually slow, check your Anthropic API rate limits — you may be hitting your plan’s throughput cap.

Have a question not covered here? Reach out at pentagon.run.