What agents can do inside Enjoy

Talking with you

• Replies. The agent can confirm it’s starting, then send one final reply when the work is done. It sends extra messages in between only when your request calls for them, such as when you ask for updates along the way.
• Conversation names. The agent gives each new conversation a short name of three to six words.
• Questions. The agent can ask up to 10 questions at once. A question can offer choices, and you can always write your own answer instead. The agent stops and waits until you answer. See Answer your agent’s questions.
• Next steps. A reply can suggest up to three follow-up requests, which you can send by selecting them. See Use suggested next steps.
• Saved changes. A reply can list the commits the agent made, which are saved points in your project’s history.

Working with your project in Enjoy

• Docs. The agent can find, read, create, and update the docs in Files. Archived docs are left out. If someone edits a doc while the agent is updating it, the agent’s change is refused rather than overwriting theirs. See Create a document.
• Workflows. The agent can read your workflows, save new ones, and change existing ones, including their schedules. Saving a workflow doesn’t run it, but a schedule is active as soon as it’s saved. Workflows installed from a collection are read-only, so the agent saves its own copy instead. See Ask your agent to create or change a workflow.
• Conversations. The agent can read the project’s other conversations, save a draft conversation for later, and start new conversations with a task. Each request can start up to five. A started conversation can use a different installed agent, model, or effort, but it never gets broader permissions than the one that started it. Started conversations work on their own.
• Links. When the agent creates something, Enjoy gives it a link, so its reply can point you straight to it.

Showing you what it built

• Previews. When a website is ready, the agent can open it beside the conversation, where you can also see it from your other devices. It can test the page by clicking, typing, scrolling, and taking screenshots. Screenshots, and some ways of typing and pointing, work only on a Mac. See How agents use the built-in browser.
• Processes. When the agent starts something that keeps running, like a website server, it registers it so it shows in Processes. Enjoy removes it from the list once it stops. See What’s in the Processes list.
• Worktrees. The agent can move its work into a worktree, a separate copy of your project on its own branch, and carry on in the same conversation. See Try ideas side by side in worktrees.

What stays with you

• Approvals. Agents ask for permission through their own approval system, following the permission mode you chose. Enjoy doesn’t add a second approval step. See Choose what your agent can do without asking.
• Untrusted content. Enjoy tells agents to treat what they read in docs, conversations, and web pages as information, not as instructions to follow.

Claude Code features that are off in Enjoy

• Timers, repeating jobs, notifications, and cloud routines. Claude Code’s timed wake-ups and repeating jobs only fire while its program stays open, and Enjoy closes it once the work is done. Its push notifications need a terminal, and its cloud routines report on claude.ai instead of in Enjoy. Enjoy turns these off. For anything on a schedule, Claude saves an Enjoy workflow instead, and results arrive as replies in Enjoy.
• Hooks. Commands you may have set up in Claude Code to run automatically at certain moments, called hooks, don’t run in Enjoy.

If your organization blocks Enjoy’s tools

Some workplaces manage Claude Code with rules that only allow approved tools. If those rules refuse the tools Enjoy adds, the conversation stops with Your organization’s Claude Code policy blocks the tools Enjoy needs. Use another agent, or ask your IT team to allow Enjoy. Switch the conversation to another agent, or ask whoever manages your computer to allow Enjoy. See Switch agents in a conversation.