The Agents area in Studio shapes how the assistant behaves everywhere: what it remembers, what it can reach, and what it is allowed to do without asking first.
Skills. Reusable instructions you write once and call in a chat with a short shortcut name, optionally limited to the tools they need.
Memory. Notes Hollow keeps across conversations. You can add them yourself, approve ones Hollow suggests, scope one to a Space, and delete any of them.
Automation rules. Checks that run around Hollow's actions: before or after a tool runs, when you send a message, or when a turn ends. They can block, ask first, rewrite input or call your webhook.
Agents. AI agents you define, each with its own system prompt, its own model choice, and its own tool groups and permission rules.
Connected tools. Connect Model Context Protocol servers to give Hollow their tools. Only servers that speak streamable HTTP at a public https address are supported.
Permissions. Build approval sets to choose which actions need your approval, and review the tools you have told Hollow to always allow, each of which expires on its own.