Hi everyone!
I’ve been thinking about building an AI assistant that works differently from what I see out there. The main idea is using OAuth2 for temporary permissions instead of storing API keys forever.
Here’s how I imagine it working:
When I tell the bot “analyze sales data and save it to my spreadsheet”, it would ask me to authorize access to my account, do the work, then immediately delete the access token. Next time I need something similar, it asks for permission again.
Same thing if I want it to create a post for social media or update my calendar. Fresh authorization each time, no permanent access stored anywhere.
What I want to build:
- Support for multiple platforms (spreadsheets, social networks, productivity apps)
- OAuth2 flows that get wiped after each task
- Clear logs showing exactly what was accessed
- Just-in-time permissions instead of always-on access
I’ve looked around but most AI tools either don’t connect to external services properly, or they’re locked into one company’s ecosystem. The ones that do integrate seem to want permanent API access.
Has anyone seen something like this? Or tried building it themselves? I’m curious if there are technical reasons why this approach isn’t more popular.
Thanks for any insights!