Visual Assist adds AI through VA Intelligence, a set of optional capabilities that extend your editing workflow and do not replace it. Navigation, refactoring, and code inspection rely on parsing and explicit rules, so you get predictable results. Use VA Intelligence to complement them with open-ended tasks, where Visual Assist can’t anticipate what you want to do.
Visual Assist AI Functions
The AI integration with Visual Assist is focused on:
- Change Code with AI: lets you describe a change in a prompt. The agent proposes the code, and you review and apply it in the VA Agent window.
making AI assistance feel more responsive, transparent, and better integrated into normal Visual Studio workflows.
Supported AI Agents
VA Intelligence supports three AI agents. Choose the one that matches the account you have and how you want your code handled:
- Codex (default): uses OpenAI’s Codex agent. Requires the Codex CLI, installed and signed in with an OpenAI account.
- Claude Code: uses Anthropic’s Claude agent. Requires the Claude Code CLI, installed and signed in with an Anthropic account.
- Ollama - gemma4 (local): runs on your computer using the Gemma model installed with VA Intelligence. No sign-in is required.
Codex and Claude Code send your code to their providers for processing. The local agent keeps your code on your computer.
Visual Assist AI Access
AI functionality is available through VA Intelligence. It is disabled by default. For security and safety, you must give permission to both install and run it.
Requirements for all agents:
- Go to VAssistX > Options > VA Intelligence.
- Enable VA Intelligence and accept the “Whole Tomato AI Providers and Restrictions on Use” terms.
- If you use the local agent, click Download Ollama & Gemma 4 (about 9 GB). The button stays disabled until you enable VA Intelligence. After the download finishes, the button reads Re-download Ollama & Gemma 4.
- If you useCodex or Claude Code, install its CLIs and sign in. See AI agent setup.
- To delete the downloaded files, click Remove Files.
Additional requirements for the local agent (Ollama - gemma4):
The local agent runs on your computer, so your hardware must support it.
- Disk space: At least 30 GB of free space
- Minimum hardware requirements: GPU with at least 32 GB of VRAM.
After you install VA Intelligence, the Change Code with AI feature is available in Visual Assist with no separate add-ons required.
Note for Upgraders:
If you used a previous version of VA Intelligence, the new install goes to %LOCALAPPDATA%\VisualAssist\ollama_bin2. The previous Ollama and Gemma 3 install in %LOCALAPPDATA%\VisualAssist\ollama_bin (about 12 GB) stays on disk. Visual Assist does not delete it, you have to delete the ollama_bin folder to free the space.
VA Agent window
The VA Agent window shows you what the agent does. When you run Change Code with AI, the agent works on your open solution: it reads and searches your code with Visual Assist tools, answers your questions, and proposes changes. The agent’s response and each proposed change appear in this window, so you can review every change before it reaches your code. To accept a change, click its Apply link.
You can run several conversations at the same time.
For more information about reviewing and applying changes in the VA Agent window, see VA Agent.
Recommended Scope
- Use AI for changes that are easy to describe but awkward to make by hand or with a fixed refactoring command.
- Keep your selection focused on the code you want to change. The agent reads surrounding code on its own, and you can review smaller changes more easily.
- Treat the output as proposed code, not automatically trusted code.