Our products · Developer tools · AI coding agents
AgentBell
Know the moment your AI coding agent needs you.
Check-ins for AI coding agents. A one-line hook sends the status of Claude Code, Codex, Cursor or any command to a live board, and pings your phone when an agent waits for an answer, asks permission, finishes or fails. It sends status only, never code.
Live at agentbell.surgesoftware.co.in · free for 1 computer · Pro $5 a month
Pricing: Free for 1 computer and a 14-day Pro trial; Pro $5 (₹449) a month, Team $6 per user a month (online payment opening soon).
Who it is for
The buyer and the problem
Developers and engineering teams who run AI coding agents on one or more computers, on Windows, macOS or Linux.
- Agents stop and wait for a yes or an answer, and nobody notices for an hour.
- Developers babysit terminal tabs instead of doing other work.
- With several computers and teammates, nobody knows which agent needs whom, or what it cost.
How it works
AgentBell, step by step
- 01
Install
Run one command. It connects the computer with a code you approve and adds Claude Code hooks.
- 02
Alerts
Turn on push on your phone (or Telegram) from the web app.
- 03
Work
Your phone buzzes when an agent waits, finishes, fails or looks stuck.
- 04
Approve
Optionally allow or deny simple permission prompts from the phone.
Features
What it does
- Live board of every agent: working, needs you, finished, failed
- Phone push and Telegram alerts
- Approve or deny Claude Code permission prompts from the phone (opt-in)
- Cost, tokens and changed file names per session
- Stuck-agent alerts and a daily digest
- Team board, share links, installable web app
Integrations
Works with
- Claude Code hooksAvailable
- OpenAI Codex CLIAvailable
- Cursor hooksAvailable
- TelegramPlanned
Under the hood
Where AI is used, and where it is not
The model
No model inside. AgentBell watches the agents you already run and reports their status.
Rules and code
Statuses come from the agents' own hook events. Cost is counted on the developer's computer from the local log, and only totals are sent.
People in the loop
Nothing is approved by AgentBell itself. A person taps Allow or Deny, and with no answer the normal terminal prompt appears.
Use cases
How businesses use AgentBell
Solo developer
Start a long refactor, walk away, and get a push when Claude needs a yes.
Several machines
Laptop, desktop and a build box on one board, with the waiting ones on top.
Team lead
A morning digest of what every agent did, with hours, cost and files changed.
Privacy-minded teams
Status only, never code. Private mode sends only status and numbers.
What it shows about Surge
The capability behind the product
AI for code and engineering
We build the tooling around coding agents that we use ourselves every day.
Privacy by design
A one-file, readable hook that keeps code and secrets on the developer's machine.
Related accelerators: Agentic IT support desk with grounded answers and human hand-off
Book a working session
A working session with our architects on one of your use cases. You bring a process and sample documents; we leave you with a scoped plan: model, hosting, integrations, timeline and cost range.