All Features

Set up new projects by connecting a GitHub repository. Configure the project name, repository URL, and access credentials.

View all your projects in a dashboard with real-time workspace status indicators showing availability, active sessions, and review state.

Update project configuration at any time: change the name, update credentials, switch runtime environments, or modify agent instructions.

Delete projects safely with soft deletion. Deleted projects move to a "Deleted Projects" section for potential recovery.

Changed your mind? Restore soft-deleted projects back to active status with a single click.

Permanently remove projects when you're certain. A confirmation dialog prevents accidental permanent deletion.

See at a glance which projects have active editing sessions, are awaiting review, or are available for new work.

Direct links from the project list to GitHub repository, workspace branch, and pull requests for fast navigation.

Describe changes in natural language: "Update the hero headline" or "Change the support email." The AI agent makes the edits for you.

Begin a new editing session for any available project. The workspace initializes automatically if needed.

Pick up where you left off. Active conversations can be resumed from the project list.

Watch the AI's response stream in real-time. See changes as they're being made, not just the final result.

AI responses are rendered as formatted markdown with code highlighting, lists, and proper typography.

Scroll through the complete conversation history. Every instruction and response is preserved.

Keep the newest messages in view automatically, or disable auto-scroll to read at your own pace.

Submit instructions by pressing Enter, or toggle off to write multi-line messages.

See "Thinking" and "Working" badges showing AI activity. Track tool call counts during active sessions.

Expand technical containers to see exactly which files were read, modified, or created.

Auto-discovered list of built HTML pages updates every 3 seconds. Click to preview your changes instantly.

The agent provides browser-accessible preview links that you can share with colleagues for review.

End your editing session when done. The workspace becomes available for others or for your next session.

Submit your changes for review with one click. A GitHub pull request is created automatically.

Access read-only views of past conversations to see what changes were discussed and made.

The agent suggests descriptive commit messages that reflect actual changes, in the language you're chatting in.

Clickable prompt suggestions from your project configuration. Hover to expand, click to insert at cursor position.

Configure external asset manifest URLs per project. Reference images and media from centralized asset repositories.

Collapsible browser panel displays all available remote assets. Search and filter to find what you need quickly.

Visual thumbnails for image assets in the browser. See what you're selecting before inserting into content.

Select an asset and click to insert its URL directly into your chat message or instruction.

Upload new images directly to S3 from the editing interface. Supports IAM role assumption for secure access.

Drop images directly into the upload area. Responsive side-by-side layout on wider screens.

Filter assets by filename or path. Regex-powered search for advanced filtering needs.

The AI agent can list, search, and get info about remote assets to help with content editing tasks.

Workspaces initialize automatically when you start your first conversation. No manual setup required.

Watch workspace initialization with a progress indicator. Status updates in real-time until setup completes.

Reset a workspace to start fresh. Ends active sessions and returns the workspace to available state.

If a workspace enters an error state, a dedicated problem page offers recovery options to reset and try again.

Color-coded badges show workspace state: available, in setup, in conversation, in review, merged, or problem.

See who's currently using a workspace. The active user's email is shown when a conversation is in progress.

10-second heartbeats track active sessions. Stale conversations are automatically released after 5 minutes of inactivity.

Each workspace runs in an isolated Docker container with resource limits. No cross-workspace access is possible.

Link any GitHub repository containing your website or landing pages. Supports both public and private repositories.

Use fine-grained or classic personal access tokens. UI guidance helps you create tokens with the right permissions.

Each workspace gets its own Git branch. Changes are isolated until merged, keeping your main branch clean.

When you send changes for review, a GitHub pull request is created automatically with all your commits.

Direct links to pull requests from the project list and reviewer dashboard. One click to see the diff on GitHub.

View the workspace branch on GitHub directly to see all committed changes before PR creation.

Jump to the GitHub repository from anywhere in SiteBuilder to view the full codebase.

Every change is tracked in Git with clear commit messages. Full history is preserved for auditing.

Dedicated dashboard for reviewers showing all workspaces awaiting review with project details and quick actions.

See the original conversation that led to the changes. Understand context before reviewing the PR.

After merging on GitHub, mark the workspace as merged. This transitions it to merged status for tracking.

Need revisions? Reopen a workspace for additional edits without losing the existing branch and PR.

Reviewer role provides access to the review dashboard and approval actions. Regular users can only submit for review.

Workspaces in review are locked. No additional edits can be made until a reviewer reopens them.

Visual progress bar shows tokens used vs. maximum context window. See input and output tokens separately.

See estimated costs as you chat. Input cost, output cost, and total are displayed to 4 decimal places.

Always know which AI model is being used for your editing session.

See how many tool calls the agent has made during the session. Helps understand complexity and cost.

Use your own LLM API keys. Pay directly to your provider with full visibility into your own usage dashboard.

Already configured a key? Select from previously used keys when setting up new projects.

Keys are validated in real-time against the provider API. Know immediately if there's an issue.

Paste your key and SiteBuilder automatically detects whether it's OpenAI or Anthropic.

Set context for the agent: what kind of project is this? What's the brand voice? What conventions should it follow?

Guide how the agent approaches tasks. Should it be conservative or creative? Ask questions or make assumptions?

Control final output format. What should responses include? How should they be structured?

Sensible defaults based on project type. Landing pages get different defaults than documentation projects.

Choose which Docker image to use: Node.js 18/20/22, Python 3.12, PHP 8.4, or specify a custom image.

Need special tools? Specify any Docker image and SiteBuilder validates it exists before saving.

Choose your AI provider. Currently supports OpenAI and Anthropic with extensible provider architecture.

Agent configuration lives in a collapsible "Advanced" section. Quick setup stays simple; power users can customize.

Place markdown files in .sitebuilder/rules/ to give the agent project-specific conventions and guidelines.

All agent commands execute in ephemeral Docker containers. Containers are destroyed after each command.

Each workspace is mounted at /workspace inside the container. The agent can only see and modify its own files.

All file operations are validated. Attempts to access paths outside the workspace (using ../) are blocked.

Containers run with defined memory (2GB) and CPU (2 cores) limits to prevent runaway processes.

Operations in one workspace cannot read or write files in another. Complete isolation between projects.

All forms include CSRF tokens. Requests without valid tokens are rejected.

GitHub tokens and API keys are stored securely. Masked input fields prevent shoulder surfing.

Users can only access their own active conversations. Role-based access controls protect sensitive operations.

Full dark theme throughout the application. Toggle between light and dark mode with persistence.

Mobile-friendly layouts that adapt to any screen size. Work from your phone, tablet, or desktop.

Spinners and progress indicators during async operations. Never wonder if something is happening.

Success, error, and warning messages displayed as toast-style alerts. Clear feedback on every action.

Advanced settings, deleted projects, and technical details live in collapsible sections. Clean UI by default.

Full keyboard support with proper focus management. Navigate efficiently without a mouse.

Helpful messages when there's no data to show. Guidance on what to do next instead of blank screens.

Dedicated error pages with recovery options. Clear explanations of what went wrong and how to fix it.

Public /workflow-docs page with interactive Mermaid diagrams showing workspace lifecycle states and transitions.

Full English and German locale support with ~160 translation keys across all templates and UI components.

Clean URL structure with /en/ and /de/ prefixes. Automatic locale detection and seamless navigation.

Toggle between languages from anywhere in the application. Your language preference persists across sessions.

All 14 application templates fully translated. Consistent experience regardless of language selection.

Sign up with email and password. Get started with your own projects immediately.

Password-based authentication with secure session management.

View your account details including email address and member-since date.

Securely end your session from any page. Session is fully cleared on sign out.