In-App Agent
Preflight includes a dedicated In-App Agent designed for security, privacy, and full compatibility with the Mac App Store. Unlike the Code Agent, which can execute local commands and requires broad system access, the In-App Agent runs entirely within Preflight’s secure sandbox.
Bring Your Own Key (BYOK)
Section titled “Bring Your Own Key (BYOK)”Because the In-App Agent runs locally on your machine and connects directly to AI providers, you must provide your own API key. Preflight does not proxy these requests or include a shared key.
Supported Providers
Section titled “Supported Providers”- OpenAI: Supports models like GPT-4o and o1.
- Anthropic: Supports Claude 3.5 Sonnet and Haiku.
Configuration Steps
Section titled “Configuration Steps”- Open Preflight’s Settings (⌘+,).
- Locate the API Keys (Bring Your Own) section.
- Paste your API key into the corresponding field (OpenAI or Anthropic).
- The key is instantly validated and stored securely in your macOS Keychain.
Capabilities
Section titled “Capabilities”The In-App Agent has access to the full suite of Preflight Tools, enabling it to:
- Inspect: Analyze USD stage hierarchy and prim properties.
- Validate: Check scenes for errors using standard validators.
- Fix: Auto-generate functionality-preserving fixes for common issues.
- Edit: Scale models, select prims, and modify scene composition.
It cannot execute arbitrary shell commands or access files outside of the documents you have explicitly opened.
Sandboxed vs. Unsandboxed
Section titled “Sandboxed vs. Unsandboxed”Depending on your version of Preflight, different agent capabilities are available.
| Feature | In-App Agent (MAS / Sandboxed) | Code Agent (Direct / Unsandboxed) |
|---|---|---|
| Availability | Mac App Store & Direct | Direct Download Only |
| Execution Environment | Sandbox (Inside App Container) | Host System (Shell Access) |
| Privacy | High (Isolated Scope) | Moderate (User Permissions) |
| Model Source | BYOK (OpenAI / Anthropic) | Local (Ollama), LM Studio, or Cloud |
| Capabilities | Preflight Tools Only | Preflight Tools + Terminal / Scripts |
| Connection | Direct to Provider | Via preflight-mcp Bridge |
Which one should I use?
Section titled “Which one should I use?”- Use In-App Agent if you want a secure, zero-setup assistant that “just works” for checking and fixing USD files, or if you are using the Mac App Store version.
- Use Code Agent if you are a power user who wants the agent to run Python scripts, use local LLMs, or integrate with external pipelines.