Getting OneDrive Working with Self-Hosted AI Agents: A Survival Story

February 18, 2026 | Last updated: February 18, 2026

Categories: story, onedrive, multcloud, automation, openclaw

This post was co-created with Clawsistant, my OpenClaw AI agent. Yes, an AI and a human brainstormed this mess together.

If you read my previous post, you know how to set up Google APIs for your AI agent. What I didn't mention? Getting Google Drive working was the easy part. The real adventure was OneDrive.

This is the story of how I finally got my AI agent to access files from OneDrive — and why I ended up using a third-party service to make it happen.

The Problem

I store a lot of important files on OneDrive: tax documents, resumes, work-related PDFs, family photos. My AI agent needed access to these files for tasks like:

• Finding and summarizing documents
• Organizing files
• Accessing work materials while working remotely

Simple enough, right? Just set up the OneDrive API like we did with Google. How hard could it be?

The Answer: Very Hard

Attempt 1: Microsoft Graph API

Microsoft provides the OneDrive API through Microsoft Graph. In theory, it's similar to Google APIs:

• Create an Azure App registration
• Configure OAuth permissions
• Get client credentials
• Connect to your agent

The reality: Microsoft's developer ecosystem feels like it was designed by a committee that never talked to each other. Here's what went wrong:

• Consent issues: Even with a personal Microsoft account, the OAuth flow kept failing with cryptic error messages
• Permission scopes are confusing: Files.Read, Files.Read.All, Files.ReadWrite — which one do I need?
• Token refresh problems: Unlike Google's relatively stable token system, Microsoft's refresh tokens have all sorts of edge cases
• No straightforward "service account" option: For personal accounts, you're stuck with delegated permissions, not the application permissions you'd use in an enterprise setup

Attempt 2: rclone

rclone is the gold standard for command-line cloud storage sync. It's free, open-source, and supports OneDrive out of the box.

The setup involves:

1. Installing rclone
2. Running rclone config to authorize OneDrive
3. Using rclone commands to sync/mount folders

The problem: rclone works great for a human typing commands, but integrating it with an AI agent? That's tricky. The agent would need to:

• Execute shell commands
• Handle the OAuth flow interactively
• Manage mounted filesystems

It's doable, but it adds complexity. I wanted something cleaner.

Attempt 3: Other Options

I also looked at:

• onedrive-docker — Docker containers that sync OneDrive locally. Good, but still requires server resources and maintenance.
• ownCloud/Nextcloud — Self-hosted alternatives. Overkill for my use case.
• Direct WebDAV — OneDrive doesn't natively support WebDAV anymore.

The Solution: MultCloud

After weeks of frustration, I discovered MultCloud — a service that lets you manage multiple cloud storage providers from one interface.

Here's why it won:

• Easy setup: Connect OneDrive and Google Drive (or other clouds) with a few clicks
• Built-in sync: Schedule automatic sync between clouds
• Agent-friendly: The AI agent talks to Google Drive (which we already set up), and files are automatically synced from OneDrive
• Reliable: No more OAuth headaches

How It Works

1. Sign up for MultCloud (free tier available, paid for more features)
2. Add OneDrive — authorize access to your OneDrive account
3. Add Google Drive — authorize access
4. Create a sync job — choose which OneDrive folders to sync to Google Drive
5. Schedule it — sync hourly, daily, or in real-time

The Trade-offs

Is MultCloud perfect? No. Here's the reality:

Pros	Cons
Simple to set up	Third-party dependency
Works reliably	Subscription cost (or one-time deal)
No OAuth headaches	Data passes through their servers
Supports 80+ clouds	Free tier has limits

For me, the trade-off was worth it. My time is valuable, and MultCloud lets me focus on actually using my AI agent rather than debugging OAuth flows.

What I'd Do Differently

If I were starting fresh today, here's my advice:

1. Start with MultCloud from day one if you need multi-cloud access
2. Don't waste time on Microsoft Graph unless you have a business need
3. Use rclone if you want a free, self-hosted solution and don't mind the CLI
4. Consider the sync approach: One-way sync (OneDrive → Google Drive) is simpler than two-way

Lessons Learned

This whole saga taught me a few things:

• Not all clouds are equal: Google's developer experience is far ahead of Microsoft's
• Workarounds are valid: Sometimes the "elegant" solution isn't worth the headache
• AI agents need stable APIs: The less friction in cloud access, the more useful the agent
• Third-party services exist for a reason: MultCloud fills a real gap

What's Next

Now that my AI agent has access to both Google Drive and OneDrive (via sync), the automation possibilities have opened up. I can:

• Search across both clouds
• Get file summaries
• Automate document organization

Next, I'm working on more advanced automation — maybe automatic file categorization or intelligent document routing. Stay tuned!