How to Decouple Your AI Workflow from Model Providers

Dylan Davis2026-06-22go watch the original →

the gist

Avoid vendor lock-in by moving your AI instructions, memory, and skills into local files on your desktop or shared drive rather than relying on native browser-based AI memory.

Externalizing AI Context

To prevent vendor lock-in, move your AI configuration from proprietary browser-based memory into local files. By using desktop agents that read from your local file system or a synced shared drive, you ensure that your instructions, preferences, and skills remain portable across different model providers. This approach treats the AI model as a replaceable utility while keeping your business logic and context in files you own.

The Three-File Setup

Organize your project folders to contain three specific types of files that define how the AI interacts with your data:

Memory Files: Extract existing AI memory by prompting the model to write out everything it knows about your preferences and work style for a specific task. Use the following prompt to export this data:
```
I am moving to a new AI tool. Write out everything that you know about me and how I work when it comes to this specific task. Pull out the preferences, style, or anything that you have picked up from me doing this task. Put it all into one block so I can easily copy it over to the new AI tool.
```
Going forward, instruct the AI to externalize new lessons into a dedicated file rather than storing them in its native memory.
Skill Files: Since most AI agents use open standards for skills, you can migrate them by opening your new tool and asking it to duplicate the skills from your previous tool's directory into the new tool's required location.
Instruction Files: Keep core instructions under 100 lines. These files should act as pointers that tell the AI the purpose of the folder and where to find relevant data. When switching tools, simply duplicate the instruction file and rename it to match the naming convention of the new agent (e.g., renaming claude.md to agents.md).

Scaling for Teams

Sync your project folders to a shared drive like SharePoint, Dropbox, or Google Drive. By keeping these files synced to your desktop, you allow multiple team members to use different AI tools while accessing the same underlying instructions and memory. This ensures that improvements made to a workflow benefit the entire team regardless of which specific model they prefer.