Skip to main content
This page walks through every step of the Whim onboarding flow in detail. If you just want to get started quickly, see the Quickstart.

​
Overview

The onboarding flow takes you from sign-up to a fully configured workspace. Here’s the full sequence:
1

Sign up and profile

2

Install GitHub App

3

Enter your name

4

Select a repository

5

Choose AI provider

6

Configure workspace

7

Enable notifications

​
Step 1: Sign up and profile

After signing in with GitHub, you’ll be asked to fill out a short profile:
  • Full name — How you’ll appear to teammates
  • Email — For notifications and account recovery
  • Occupation — Helps Whim tailor the experience
  • Team size — Whether you’re working alone, in a small team, or at a larger organization
This information is collected once and can be updated later in your account settings.

​
Step 2: Install the GitHub App

Whim needs access to your repositories via the Whim GitHub App. If you don’t have it installed yet, you’ll see a prompt to install it. Clicking Install GitHub App redirects you to GitHub where you can:
  • Choose your personal account or an organization
  • Grant access to all repositories or select specific ones
Once installed, Whim links the installation to your team and fetches your repository list.
If your organization already has the Whim GitHub App installed, this step is skipped automatically. Whim detects existing installations and moves you forward.

​
Step 3: Enter your name

If Whim doesn’t have your first name on file yet, you’ll be asked to enter it. This is used for display purposes in the workspace — task attribution, comments, and collaboration features.

​
Step 4: Select a repository

You’ll see a list of repositories accessible through your GitHub App installation. Select the repo you want to create a workspace for. Each repository shows:
  • Name and owner
  • Primary language (if detected)
  • Whether it’s already connected to an existing workspace
Creating a new repo: You can also create a brand-new GitHub repository directly from this screen. Enter a name, optional description, and choose public or private visibility. Whim creates the repo on GitHub and immediately sets up the workspace.
Once you select a repository, Whim creates the workspace and syncs the code. This typically takes a few seconds.

​
Step 5: Choose AI provider

After workspace creation, you’ll configure your AI provider. This is where you decide which AI model powers your tasks.

​
Available providers

The default provider. Claude Code runs natively in each task container.Connection options:
  • Whim compute — Use Whim’s built-in CU allocation. No API key needed.
  • Claude subscription — Bring your existing Claude Pro or Team subscription by logging in through the provider setup.
  • API key — Enter your own Anthropic API key for direct access.
OpenAI’s Codex CLI runs natively in task containers.Connection: Requires an OpenAI API key. Enter it during provider setup.
Access 10+ models from multiple providers through a single integration.Connection: Requires an OpenRouter API key. Once connected, you can choose from models like GPT-4o, Gemini, Grok, DeepSeek, and more on a per-task basis.
You can configure multiple providers and switch between them at any time. The provider you set here becomes the default for new tasks in this workspace.

​
Step 6: Configure workspace

After choosing your provider, you’ll enter the workspace configuration wizard. This is where you fine-tune how your AI agents behave.

​
Setup options

At the intro screen, you have three choices:
  1. Set up manually — Walk through each configuration step
  2. Import from local setup — Pull your existing Claude Code/Codex configuration from your local machine
  3. Skip setup entirely — Use defaults and configure later

​
Manual setup steps

If you choose manual setup, you’ll walk through these steps (all are optional — skip any you don’t need):
Choose the default AI model for this provider. Options vary by provider:
  • Claude: Opus, Sonnet, Haiku (with fast mode toggle)
  • Codex: Available OpenAI models with configurable reasoning effort (Low, Medium, High, X-High)
  • OpenRouter: Select from 10+ models across providers
Control how much autonomy the AI agent has in task containers.Claude permission modes:
ModeBehavior
DefaultPrompts for permission on first use of each tool
Accept EditsAutomatically accepts file edit permissions
Plan ModeCan analyze but not modify files
Bypass PermissionsSkips all permission prompts
Codex permission modes:
ModeBehavior
DefaultUses Codex CLI defaults
UntrustedOnly trusted commands run without approval
On RequestRequests approval when needed
On FailureAsks only when a command fails
Never AskNo approval prompts (sandbox still applies)
Danger Full AccessNo approvals and no sandbox protections
Write custom instructions that the AI agent will follow in every task. This is equivalent to a CLAUDE.md file — project-specific guidelines, coding standards, or context the agent should know.Supports Markdown formatting with a live preview toggle.
Define environment variables that will be injected into task containers. Useful for:
  • API keys and secrets
  • Feature flags
  • Service URLs
Variables are stored securely and organized by file (e.g., .env).
Configure Model Context Protocol servers that extend the AI agent’s capabilities. MCP servers give agents access to external tools and data sources.
Enable or disable plugins that add functionality to your workspace. Plugins can extend the agent’s toolset, add integrations, or customize behavior.
Configure custom skills (slash commands) that agents can use. Skills are reusable prompt templates that can be invoked with /<skill-name> during a task.
Define shell scripts that run in task containers:
  • Init script — Runs when a task container starts (install dependencies, set up tools)
  • Bashrc — Appended to the container’s .bashrc (aliases, PATH modifications)
Configure how long a task container stays running after the agent goes idle. Options range from 1 minute to 1 hour, or disabled entirely.The default is 15 minutes. Sleeping tasks resume instantly when you interact with them.

​
Import from local setup

If you already have Claude Code or Codex configured locally, you can import that configuration:
  1. Whim generates a short-lived import token
  2. Run the provided CLI command on your local machine
  3. Whim detects the import and pulls in your local settings — model, permissions, MCP servers, plugins, skills, env vars, and scripts
  4. Review the imported config and adjust as needed
The import token expires after a few minutes. If it expires, you can generate a new one from the same screen.

​
Save as defaults

After completing configuration, you’ll have the option to:
  • Set as workspace defaults — Apply this config to all new tasks in the workspace (if you’re a workspace admin)
  • Save as your personal defaults — Use this config as your starting point when joining other workspaces

​
Step 7: Enable notifications

The final step is enabling browser notifications. Whim can notify you when:
  • An AI agent finishes working and goes idle
  • Someone mentions you in a comment
  • New comments appear on your tasks
You can enable notifications now or skip and configure them later in settings.
After completing onboarding, you’re taken to your workspace where you can start creating tasks. See Creating and Running Tasks for next steps.