Wallfacer
On this page

Quickstart: App

From zero to a working AI coding session.

Before You Begin: The Waiting Room

If your account is still being activated, the app shows a waiting room instead of the full workspace. Once your account is ready, the workspace opens automatically.

If you landed here on the wrong account, use the Sign out link at the bottom of the waiting room, then sign back in with the account you meant to use.

Prerequisites

  • A GitHub account (this quickstart builds on a code repository; see GitHub Integration for when GitHub is optional)
  • A Wallfacer account
  • For the iPhone or iPad app, iOS 27 or iPadOS 27 or later. Devices on version 26 cannot install the current app. You can use Wallfacer in a browser instead.

Step 1: Connect Your Agent to GitHub

This quickstart clones a repo, so GitHub is required here. Work that doesn't use a repository can skip this step.

Your account's agent holds the GitHub identity that clones repos and opens pull requests. Connect it once (this is admin or owner work):

  1. Go to Agents and open your account's agent, or use the Connect your agent to GitHub onboarding step.
  2. In Credentials > GitHub, paste a token for the GitHub account the agent should act as. Use a dedicated bot account so its work is clearly attributable.
  3. Tap Connect. The section then shows the account the agent is connected as.

See GitHub Integration for details.

Step 2: Create an Agent

You assign work to an agent -- a configured worker with its own computer, model, and identity. Create one before you start coding.

  1. Open Agents and choose New agent
  2. Pick a role template and give the agent a name
  3. Wallfacer creates the agent and provisions its computer (an environment for the repository you'll work on). The computer installs dependencies and starts services -- this may take a few minutes the first time.
  4. On the agent's page, add a model credential so it can start working, and connect a GitHub identity if it needs to work with your repositories

See Agents for the full model, and Environments for what a computer contains.

Step 3: Create a Task

  1. Click New Chat and pick your agent. Its computer and model are set for you automatically.
  2. You land in the task screen, empty, with the chat input at the bottom.
  3. Type a message describing what you want built and send it. Sending it is what creates the task.

Step 4: Start Coding

A VM boots from your snapshot. You're in a live chat with Claude Code.

  • Send messages to direct Claude's work
  • Watch file changes appear in the diff viewer as Claude writes code
  • Claude auto-commits to a branch as it works

Step 5: Review

When files change, the pill above the task composer shows the changed-file count and lines added and removed. Select it to review the diff. On phones, the Code Changes tray lets you switch between the Code diff and a Files list that covers every repository in the task. On tablets and desktop, the pill opens the Diff pane. Additions appear in green and deletions in red. See Code Review for file navigation and pull requests.

What's Next

  • Agents -- How agents carry their computer, model, and identity
  • Environments -- Customize your setup commands, services, and platform
  • Sessions -- How coding sessions work
  • API Quickstart -- Programmatic access via the REST API