trigr wakes your AI agent on schedules, events, and messages — so it works while you're away.
AI coding agents can read your entire codebase, write code, run tests, send emails, and more. But they only work when you sit down and type. Close the laptop, and the agent goes dark.
Your agent sits idle until you prompt it. You are the bottleneck. Nothing happens while you're away, asleep, or on the train.
Your agent watches for events. A deploy fails, an issue opens, a scheduled time hits — it wakes up, handles the task with full context, and goes back to sleep.
Three simple concepts. That's the whole system.
Send your agent a task from anywhere. Another terminal, a script, a webhook. Like texting a coworker who knows your entire codebase.
trigr emit "Triage the new issue"
Cron jobs that wake the agent. Daily summaries, weekly reports, periodic maintenance — on autopilot.
trigr add standup \
--cron "0 9 * * *" \
-m "Run the daily standup"
Watch for changes at regular intervals. If a command produces output, it becomes an event. Silence means nothing new.
trigr add inbox \
--interval 300 \
-c "./check_email.sh"
$ uvx trigr watch
$ uvx trigr emit "Check if the deploy succeeded"
$ pip install trigr # to install permanently
| Command | What it does |
|---|---|
| trigr watch [--timeout 300] | Block until an event arrives, print it, exit |
| trigr emit "msg" [--delay 10s] | Send a message to the queue |
| trigr add <name> --interval N | Add a poller that runs every N seconds |
| trigr add <name> --cron "..." | Add a cron-scheduled trigger |
| trigr remove <name> | Remove a poller or cron job |
| trigr status | Show server state and active triggers |
| trigr init | Create trigr.toml in current directory |
| trigr serve [-f] | Start server manually (auto-starts on watch/emit) |
Runs as a background task. Chat while waiting.
Runs as a blocking call.
Runs as a blocking call.