pinged.fyi vs ntfy — which push notification service should you use?

Both tools solve the same beautiful problem: anything that can make an HTTP request can buzz your phone. Here's an honest comparison — including the cases where you should pick ntfy.

The 10-second version

What's the same

The core loop is identical, and it's the right loop:

curl -d "backup finished" https://pinged.fyi/t/<your-topic>

Both deliver over APNs/FCM to a phone app, both support priorities and titles, both let you subscribe from the command line.

Where pinged.fyi is different

Topics are unguessable by default. ntfy topics are names you pick — and if you pick mytopic, anyone who guesses it can spam you or read your messages. pinged.fyi generates four random words (~52 bits) for every topic, so privacy is the default, not a discipline.

Agents are first-class. The fyi CLI doubles as an MCP server. One line — claude mcp add fyi -- fyi mcp — and Claude Code can ping your phone when a long refactor finishes or a decision is needed. There's also llms.txt, so an agent told to "set up pinged.fyi" can do the entire thing unassisted.

The hosted service is just free. No message quotas to think about, no reserved-topic upsell. One person runs it on a small server; the rate limits (60 messages/topic/hour) exist to stop abuse, not to sell you a tier.

Where ntfy is the better choice

You want to self-host. ntfy is open source and excellent at this. pinged.fyi is a hosted service (self-hosting may come later).

You need web push or a web UI. ntfy has both; pinged.fyi's clients are the mobile apps, the CLI and raw HTTP.

You depend on a long ecosystem tail — ntfy has years of integrations, tutorials and community answers. pinged.fyi is young and moves fast.

Bottom line

Same idea, different philosophies: ntfy is the open-source Swiss army knife; pinged.fyi is the zero-setup hosted path with agents built in. If you're already telling Claude what to do all day, fyi is the shortest distance between "the build finished" and your pocket.