Quickstart | Creor

Go from install to your first useful code change in five minutes. This guide walks through the core Creor workflow: open a project, chat with the agent, review its changes, and use plan mode for complex tasks.

1. Open a Project

Open any project folder in Creor. You can do this from the menu (File > Open Folder), by dragging a folder onto the window, or from the terminal.

creor ~/projects/my-app

Creor works best when you open the root of a project — the directory that contains your package.json, Cargo.toml, go.mod, or similar project file. This gives the AI agent the full context of your codebase.

Tip

If your project has a monorepo structure, open the monorepo root. The agent understands workspaces and can navigate between packages.

2. Start a Chat

Press Cmd + L (macOS) or Ctrl + L (Windows/Linux) to open the AI chat panel. This is your primary interface for talking to the Creor agent.

The chat panel opens on the right side of the editor. You can resize it, move it to the bottom, or pop it out into its own window. The agent has access to your entire project — it can read files, search code, run terminal commands, and make edits.

Type a message and press Enter to send it. The agent will respond with a plan, ask clarifying questions, or start working immediately, depending on the complexity of your request.

3. Make Your First Edit

Try asking the agent to make a real change to your code. Here are some good first prompts to try:

Add a dark mode toggle to the settings page

Write unit tests for the auth middleware

Refactor the UserCard component to use TypeScript generics

Fix the bug where the sidebar doesn't close on mobile

The agent will read the relevant files, figure out the right approach, and apply changes directly to your code. You will see each step in the chat — file reads, searches, edits, and terminal commands — as the agent works through the task.

Permissions

By default, the agent asks for permission before running shell commands or making edits. You will see a permission card in the chat with "Allow" and "Deny" buttons. This keeps you in control while the agent works.

You can adjust permission levels in Settings > AI > Permissions. Options range from "Ask for everything" to "Auto-approve read-only operations" to "Auto-approve all" for trusted projects.

4. Review Changes

After the agent makes edits, Creor shows the changes in a built-in diff viewer. You can review every modification before accepting it.

Inline diff

Changed files appear with color-coded gutters in the editor — green for additions, red for deletions. Click any changed region to see the before/after side by side.

Diff viewer panel

For a full overview, open the diff viewer from the chat panel. It shows all files the agent modified in the current session, grouped by change type. You can accept or reject individual changes, or accept all at once.

Undo

Every agent action is tracked. You can undo any change by clicking the revert button on the specific tool call in the chat history, or use the standard undo (Cmd + Z) in the editor. The agent also supports session- level revert, which rolls back all changes from a conversation turn.

5. Plan Mode

For complex, multi-step tasks, use Plan Mode. The agent will create a detailed plan before writing any code, giving you a chance to review and adjust the approach.

Activate plan mode

Press Shift + Cmd + P (macOS) or Shift + Ctrl + P (Windows/Linux) to toggle plan mode on. You will see a "Plan" badge in the chat input area. Alternatively, start your message with "/plan".

How it works

You describe what you want to build or change.
The agent reads your codebase, explores the relevant files, and creates a step-by-step plan.
The plan appears as a structured checklist in the Plan panel. Each step shows which files will be touched and what will change.
You review the plan, suggest modifications, or approve it.
Once approved, the agent executes each step, checking off items as it goes.

When to use plan mode

Plan mode is ideal for tasks that touch multiple files, require architectural decisions, or where you want to verify the approach before any code is written. Examples:

Adding a new feature that spans frontend and backend.
Refactoring a module with many dependents.
Setting up CI/CD, testing infrastructure, or database migrations.
Any task where you would normally write a design doc first.

Tips for Effective Prompting

Be specific about the outcome

Instead of "improve the login page", try "add email validation to the login form that shows an inline error message below the email field when the format is invalid". The more precise your request, the better the result.

Point to files and functions

The agent can find things on its own, but you will get faster results if you mention specific files or functions. You can reference files with @-mentions in the chat — type @ and start typing a filename.

@src/components/LoginForm.tsx add email validation with a regex check

Iterate in the same session

The agent remembers context within a session. After it makes a change, you can follow up with refinements like "also handle the case where the email field is empty" or "add a test for that validation". You do not need to re-explain the full context.

Use the right mode

For quick, focused changes — normal chat mode. For complex, multi-file features — plan mode. For parallel work — use background agents (check the Agent docs for details). Matching the mode to the task makes the agent significantly more effective.

Provide constraints

If you have preferences about how something should be built, say so upfront. "Use Zod for validation, not Yup" or "keep it under 50 lines" or "follow the pattern in UserCard.tsx". The agent respects explicit constraints.

Tip

You can define persistent rules for your project by creating a CREOR.md or .creor/rules file in your project root. The agent reads these on every session and follows them automatically. See the Rules documentation for details.