Walk away mid-run
Start the run, go make coffee. Your phone stays quiet until the agent actually needs you.
Push notifications from your coding agents
Stop watching.They’ll ping you.
Kick off a long agent run and close the laptop. You’ll get a push when it finishes, gets stuck, or needs an answer from you.
Works with the agents you already run. If it speaks MCP, it can ping you.
Codex
Claude Code
Gemini CLI
OpenCode
Copilot
Pi
A 20-minute run needs you for 5 seconds
The trick is being there for the five seconds without sitting through the twenty minutes. That’s the whole product.
One card, not thirty alerts
Live checklists update in place, so a six-step deploy is still just one card on your phone.
Unblock it from anywhere
A blocked agent asks, you tap an answer on the notification, and it picks up right where it stopped.
Every run gets its own thread
The agent opens a thread, and everything from that run stays in it: messages, checklists, questions, results, plus agent, model, and machine context when it’s sent.
You only get pinged when it matters
Progress ticks along quietly in one updating card. When something needs a decision, it comes through at the right priority with context attached.
A checklist that checks itself off
The agent posts its plan, ticks steps off as it goes, and can pin a short result note to each finished step.
Runningjust now
Ship checkout v2
Rolling out the new Stripe checkout
3 / 5
- 1.Audit billing schema3 tables touched, nothing breaking
- 2.Add webhook handlers12 events covered
- 3.Deploy checkout previewpreview-4821 is live
- 4.Run checkout integration suite84 passed, 0 flaky
- 5.Request production approvalpinged your phone
“Urgent” actually means urgent
Agents tag every ping low, normal, high, or urgent, so routine progress never sounds like a production incident.
lownormalhighurgent
Answer from the lock screen
When an agent hits a decision it can’t make, it asks. You tap “Deploy now” or “Hold” and it continues with your answer.
Setup ends with a real ping
No blind “you’re all set” screen. The wizard isn’t done until a real test notification shows up on your device.
01 · agent
Pick your agent
Choose Claude Code, Codex, Gemini CLI, or whatever you run, and get the exact MCP config for it.
02 · key
Paste one config
It comes with a dedicated API key, so each connection stays isolated and revocable on its own.
03 · ping
Wait for the ping
Send a test ping from the agent. When it lands on your phone, you know the whole path works.
The boring but important part
You’re putting a tool between you and your agents, so here’s exactly what it touches and what happens if we ever disappear.
It never sees your repo
pingmigo.app doesn’t connect to or scan your code. It stores only what your agent explicitly sends it.
One key per agent
Each key is shown once, stored as a SHA-256 hash, and revocable on its own. Cutting one agent off doesn’t touch the rest.
Hooks catch forgetful agents
Claude Code lifecycle hooks fire finished and needs-input alerts even if the agent never calls the tool.
An exit plan in writing
If the hosted service is ever discontinued, every Lifetime customer gets the source code.
Cheap enough to not think about
pingmigo.app doesn’t run your models, it just keeps them in touch with you. So it’s priced like a messaging bill, not an AI bill.
Basic
$1.99$1.39/ month
billed monthly$16.68 billed once a year
Plenty for one person running an agent or two through the workday.
- 10,000 messages a month
- Live progress and checklists
- Questions, answers, and results
- Full thread history
Tokenmaxxer
10× volume$5.99$4.19/ month
billed monthly$50.28 billed once a year
For parallel runs and agents that work longer hours than you do.
- 100,000 messages a month
- Everything in Basic
- Room for a whole fleet of agents
- 10× Basic for 3× the price
Lifetime
$24.99once
Pay once, keep it forever, and never see this line on a bank statement again.
- 10,000 messages a month, forever
- One payment of $24.99
- No renewal, no subscription
- Source code if we shut down
Questions we’d ask too
The stuff worth checking before you wire a new tool into your setup.
Is this another AI agent I have to babysit?
No. pingmigo.app doesn't run models or write code. Your agents connect over MCP and use it to report progress, ask questions, and deliver results.
Which agents does it actually work with?
Guided setup covers Claude Code, Codex, Gemini CLI, OpenCode, GitHub Copilot, Amp, and Cursor. Anything else that speaks MCP can connect with the standard config.
Does it need access to my repo?
No, and it has none. pingmigo.app never connects to or scans a repository. It stores only what the agent sends: thread, message, and activity content, plus optional agent, model, and host info.
What happens when the agent just forgets to ping me?
The installable skill teaches agents the full notification workflow, so in practice they remember. Claude Code users can also enable lifecycle hooks that fire finished and needs-input alerts even when the model never calls the tool.
Won't live progress updates spam my phone?
No. A live activity is one card that updates in place from start to finish. You also choose per device which notification types get through.
What exactly does Lifetime get me?
One payment of $24.99 for 10,000 messages a month, forever. If the hosted service is ever discontinued, Lifetime customers get the source code.
Start the next run and walk away
Your agents know how to reach you now. Pocket your phone and go do something else. It’ll buzz when there’s a decision to make or a result worth reading.
Connect your first agent