Set up with an AI agent.
Copy one prompt into your AI coding agent while it is open on your project. It reads your code first, asks only what it cannot work out, builds the integration, and sends a real test notification before it says it is done.
Plain text, ready to copy. Works with Claude Code, Codex, Cursor, Copilot, ChatGPT or similar.
What the agent does.
Looks before it asks
It reads your project first and skips any question the code already answers.
Builds it your way
A small send function in your project's own language. A failed notification never breaks the job it reports on.
Proves it works
It checks the key, sends one real test called "Nowlark is connected", and tries the failure path on purpose.
Safe to hand over.
The key can only send
An API key cannot read your history, see your devices or touch your account. Revoke it in the app at any time.
Kept out of your code
The prompt tells the agent to keep the key out of source code, logs and commits.
A mistake is capped
Each key sends at most 1,000 an hour, so a loop in generated code is stopped at that number.
Three steps from you.
1. Make a key
In Nowlark on your iPhone, open Settings, then Connect an App, and choose Create Your Key. It is shown once.
2. Put it on the machine
Set it as the environment variable NOWLARK_API_KEY in the shell or .env file the agent runs in, rather than pasting it into the chat.
3. Paste the prompt
Give your agent the prompt and answer its questions. Then check that the test notification arrives on your phone.