> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/ai-workbench/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/ai-workbench/_mcp/server.

> Technical reference for the AI Workbench build assistant including

# Build Assistant Reference

## Overview

**The build assistant is an internal AI Workbench agent that analyzes and fixes failed container builds.**

It operates within restricted file access and uses NVIDIA Endpoints for inference.
This page documents the assistant's technical constraints, file access model, and model configuration behavior.

## File Access


**Read access is limited to environment configuration files.**


**Write access is limited to build scripts.**


**Package changes go through managed tools.**


## Model Configuration

**The model catalog is hosted and updated independently of AI Workbench releases.**


**Desktop App and CLI model settings are managed separately.**


**Model removal triggers a 410 error and automatic fallback.**


## Build Warning Modal

**A review gate appears after the build assistant edits files.**


## Related Topics

- [How to Use the AI Workbench Build Assistant](/ai-workbench/how-to/ai-workbench-build-assistant) - How to use the build assistant
- [Connect the NVIDIA Endpoints Integration](/ai-workbench/how-to/integrations/nvidia-integrations#connect-the-nvidia-endpoints-integration) - Configure the NVIDIA Endpoints integration
- [Sandboxing Agents with AI Workbench](/ai-workbench/concepts/sandboxing-concept) - How agent sandboxing works in AI Workbench
- [Sandboxing Agents with AI Workbench](/ai-workbench/concepts/sandboxing-concept) - How agent sandboxing works