Get a push notification when your AI agent finishes a task

The problem with long agent runs

Modern coding agents are fast at typing and slow at finishing. A single instruction can spin off ten minutes of tool calls: reading files, running the test suite, fixing what broke, running it again. That's an awkward amount of time — too long to keep staring at the scrollback, too short to confidently start something else. So you half-watch, context-switch badly, and get nothing done in either place.

Lert closes the loop. You give the agent one job: at the very end of the run, send a push to your phone. Now you can genuinely leave. When the notification buzzes, you come back — and only then.

The one-line version

Lert gives you a unique URL. Anything that can make an HTTP request can POST to it, and your phone gets a notification. Have the agent run this as the last command of its task:

curl -X POST "https://lert.dev/p/your-id" \
  -H "Content-Type: text/plain" \
  --data-binary "Agent finished: $(git log -1 --format=%s)"

That sends the subject line of the last commit straight to your lock screen — so the notification isn't just "done," it's what got done. We use text/plain and --data-binary here because commit messages often contain =, &, or quotes that would otherwise be mangled. No headers beyond content type, no auth token, no SDK.

Wire it into your agent

1. Copy your Lert URL

Open the Lert app and your unique endpoint is on screen — something like https://lert.dev/p/your-id. Tap Copy. Treat it like a password: anyone who has it can send you a notification, and you can regenerate it any time.

2. Tell the agent to notify you

Add a line to your project's instructions (a CLAUDE.md, a Cursor rule, or your system prompt): "When you finish a task or get blocked, run curl -X POST https://lert.dev/p/your-id --data-binary '<short status>'." Because it's just a shell command, every agent that can run a terminal already knows how to do this. Send a distinct message for each state so your lock screen tells the whole story at a glance:

# run the task, then notify — success or failure
if npm test; then
  curl -X POST "https://lert.dev/p/your-id" \
    --data-binary "✅ Agent done — tests green: $(git log -1 --format=%s)"
else
  curl -X POST "https://lert.dev/p/your-id" \
    --data-binary "❌ Agent stopped — tests failing, needs you"
fi

3. Start the task and leave

Kick it off and put your phone in your pocket. You get a real iOS push — it wakes a locked phone and buzzes a paired Apple Watch, usually in under a second. No more thumb-refreshing the terminal.

Good things to send

• The finished summary: the last commit subject, the number of files changed, or a one-line result the agent wrote itself.
• A blocked signal: when the agent hits a decision it can't make — a migration it won't run, a secret it needs — so you can step in instead of finding it stalled an hour later.
• Test / lint status: green vs red is the single most useful thing to know before you decide whether to look.

Why Lert for agent alerts

Most notification tools want an account, an API key, and a client library before you can send one message — which is exactly the kind of friction that makes you not bother. Lert removes all of it: the endpoint URL is the authentication, so notifying yourself is a single HTTP request an agent can emit with zero setup. It's the difference between babysitting a run and trusting it to tap you when it matters.