We shipped a meaningful improvement to how Strug Works handles mission approvals: real-time Slack notifications and automatic escalation reminders. If you're running autonomous agents in production, you know that approval latency is one of the biggest workflow killers. This update addresses that directly.
What Changed
The alerter service—our notification orchestrator—now integrates with Slack to push approval requests directly to your team channels. When a mission reaches the awaiting_approval state, stakeholders receive an immediate alert with mission context and a link back to Strug Central. No more dashboard polling. No more missed approvals sitting in a queue.
We also added escalation reminders. If an approval request sits untouched for a configurable threshold (default: 30 minutes), the system sends a follow-up. This prevents missions from stalling when approvers are in back-to-back meetings or buried in Slack threads.
The implementation includes comprehensive test coverage (test_alerter_approval.py) to ensure alerts fire reliably under different mission states and retry scenarios. We also updated the task runner to trigger alerts at the right lifecycle hooks.
Why It Matters
Autonomous agents move fast. When Strug Works generates a mission plan and pauses for human approval, every minute of delay compounds. Engineering teams live in Slack, not dashboards. By meeting approvers where they already are, we're cutting latency from minutes (or hours) down to seconds.
This is part of a broader shift in how we think about human-agent collaboration. The best orchestration platforms don't force humans to adapt to agent workflows—they adapt agent workflows to how humans actually work. Slack is where decisions happen. Now approvals happen there too.
What's Next
We're working on making these alerts interactive. Imagine approving or rejecting a mission directly from Slack with a button click, no context switch required. We're also exploring configurable escalation chains—routing to different stakeholders based on mission risk or domain.
Longer term, we're building smarter notification routing. Not every approver needs every alert. The system should learn which humans care about which types of missions and route accordingly. That's the kind of intelligence an orchestration platform should have.
If you're using Strug Works and want to configure Slack alerts for your workspace, the setup is straightforward—check your .env.example for the required webhook URL and channel configuration. We'll be documenting best practices for alert thresholds and escalation timing as more teams adopt this.