
Looking for a proactive, cross-device AI agent. Hitting the limits of my current setup - any ideas?
Hey everyone,
I'm trying to build or find an AI assistant workflow that actually feels like a proactive assistant rather than just a chatbot I have to manually trigger. I currently have Gemini Advanced and Copilot Pro, and I'm happy to pay for a premium service if it can deliver the seamless experience I'm looking for.
My Current Setup: Right now, I'm running a local AI agent (Antigravity) on my PC connected to my personal NotebookLM via MCP (Model Context Protocol). I use Windows built-in voice typing to write.
The Good:
- Full local file access to my PC.
- Good Google Workspace API integrations.
- Built-in browser access via the agent.
The kind of things I want to automate:
- Everyday admin: "Read my latest email, find a suitable photo in my local photos folder, and reply," or taking over form-filling and file organization.
- Life logistics: Flight comparisons, event planning, and fuzzy memory recall ("I searched for a fantasy book a few weeks ago, please figure out which one it was").
- Academic/Research workflow: I'm a math researcher. I want the agent to summarize new papers, search for postdoc openings, or help refine my CV using my local notes.
- OS actions: Troubleshooting PC bugs or installing software directly.
The Friction Points (Where my setup fails):
- No Mobile Continuity: If I start an agent session on my PC, I can't just pick up my phone and continue the same conversation/session on the go. I tried syncing my local agent files to Google Drive so my phone's Gemini app could read them, but it constantly fails the read/write workflow.
- Not Proactive (No background scheduling): I want my AI to run tasks automatically on a schedule. For example: "Search for new math papers every week, give me a summary, and ask for my feedback." Or simply reminding me to cancel a hotel reservation in a week. Right now, it can't wake itself up to perform background tasks.
- Weak Autocomplete/Context: I have a massive folder of local notes, but when I'm writing new math notes, the AI doesn't actively use my repository to provide a highly personalized autocomplete.
- True "Agentic" Balance: I want an agent that just executes (e.g. runs the terminal command itself rather than giving me a tutorial on how to run it), but knows to pause and ask for feedback when data is missing.
Does anyone know of a paid service, custom platform, or API-based workflow that bridges this gap between deep local PC agents and mobile continuity while allowing proactive scheduling?