Building an AI Agent Desktop App with PowerShell, WinForms, and Multi-Provider AI

I've been working on a new AI Agent App for AdvantageBuilder and wanted to share some progress.

The goal wasn't to build another chatbot.
Instead, I wanted a desktop application capable of executing real-world tasks through a tool-driven AI agent framework.

What makes it different?

The application is built around a goal-oriented AI Agent architecture:

Goal
→ Agent
→ Tools
→ Result

Instead of generating text, the agent can decide which tools to use in order to complete a task.

Examples include:

• Reading and writing files
• Querying databases
• Calling REST APIs
• Executing PowerShell scripts
• Executing JavaScript
• Web searching
• Text-to-speech
• Delegating work to sub-agents
• Email automation

Multi-Provider Support

The AI Agent App supports multiple AI providers:

• OpenAI
• Anthropic
• Google
• Groq
• Mistral
• Ollama

Users can select both provider and model directly from the UI.

Tool Sandboxing

One feature I particularly like is the ability to choose exactly which tools are available to the agent.

For example:

✓ Gmail
✓ File Operations
✓ Database Access
✗ PowerShell Execution

This gives much finer control over what an agent is allowed to do.

Gmail Integration

One of the bigger additions was a Gmail toolset built on top of the Gmail API.

Current capabilities include:

• gmail_search
• gmail_read
• gmail_get_message
• gmail_download_attachments
• gmail_send
• gmail_reply
• gmail_delete

This allows the agent to perform workflows such as:

Search emails
→ Read email
→ Download attachments
→ Reply
→ Delete

all through tool calls.

Architecture

The project is intentionally layered:

AI Agent App
↓
AI Agent Framework
↓
Tool Layer
↓
Macros
↓
PowerShell Modules

The advantage of this approach is that adding new capabilities usually means adding a new tool rather than changing the agent itself.

Lessons Learned

A few interesting takeaways:

• Goal-oriented agents and chat applications are different products.
• Tool design matters more than prompt engineering.
• Small validation mistakes can waste hours of debugging.
• Multi-provider support is easier when abstracted from the start.
• Email automation becomes extremely powerful once combined with AI tooling.

What’s Next?

Potential future additions include:

• Microsoft Graph / Outlook integration
• Calendar tools
• Contacts tools
• OneDrive integration
• Tool activity viewer
• Saved tool profiles
• Settings persistence

The AI Agent App has become one of the best demonstrations of what AdvantageBuilder can actually do when scripting, automation, integrations, and AI are combined into a single workflow engine.
I'd be interested to hear how others are approaching tool-enabled agents versus traditional chatbot architectures.

Try AdvantageBuilder

You can download AdvantageBuilder — including the macro engine, WinFormsHelper, ObjectStore, database wrappers, and full documentation — from:

https://advbuilder.dev/

🚀 Download AdvantageBuilder