Orchestrate AI Agents from Your Phone with Mobile Approval Workflows
Orchestrate AI Agents from Your Phone with Mobile Approval Workflows
The biggest pain point with autonomous AI agents is not getting them to work - it is getting them to stop at the right moment and ask for permission. You set up an agent to handle a complex workflow, leave your desk, and come back to find it either stalled waiting for approval or went ahead and did something you did not want.
The Mobile Approval Problem
Most agent frameworks assume you are sitting at your computer watching the agent work. But the whole point of automation is freeing you to do other things. You want to be at the grocery store while your agent deploys code, drafts emails, or processes data - and only get pulled in when a decision is needed.
Webhooks with Push Notifications
The solution is straightforward: when the agent hits a decision point, it fires a webhook that sends a push notification to your phone. The notification has two buttons - Approve and Deny. You tap one, the webhook responds, and the agent continues or stops.
This pattern works because most agent decisions are binary. "Should I deploy this?" Yes or no. "Should I send this email?" Yes or no. "This file looks different than expected - proceed?" Yes or no.
What This Looks Like in Practice
Your desktop agent is running a multi-step workflow. It reaches a step that requires human judgment - maybe it is about to modify a production database or send a message to a client. Instead of blocking and waiting for you to check your laptop, it sends a notification to your phone with context about what it wants to do.
You glance at your phone, read the summary, tap Approve, and go back to your groceries. The agent continues immediately.
Building the Pipeline
The components are simple: a webhook endpoint, a push notification service like Firebase Cloud Messaging or Apple Push Notifications, and a small mobile app or shortcut that handles the approve/deny response. The agent framework just needs to call the webhook URL at decision points.
Fazm is an open source macOS AI agent. Open source on GitHub.