Mission control for AI agents
A team of agents. Every move checked.
Run several agents at once, or one lead agent with helpers. They get on with your task straight away, you watch every step, anything you have switched off is refused, and Stop works at any moment.
- Research: Working
- Flights: Waiting
- Inbox: Blocked
- Shopping: Finished
- Compare: Idle
Ask in plain words
Type a task. Sentinel picks the right agent for it.
Watch every step
A live feed names each page opened, click and app action.
Stop any time
Anything you have switched off is refused. Stop ends it at once.
Live defence, every step
Watch Sentinel think before it acts.
01 · Check
Every action passes your rules
Before any click, read or message, the step is checked against the agent's skills and your limits.
02 · Block
Hidden instructions are caught
Text on a page that tries to give the agent orders is spotted and withheld from the agent.
03 · Connect
Your apps, behind one shield
Everything read from connected apps goes through the same checks as web pages.
Why it is different
Your agents do what you ask, and stop the moment you say.
Sentinel is a desktop app. You describe a task in plain language, the agent works through it in a browser window you can see, and every step is checked against your own limits before it happens.
You see every step
A live feed names what the agent opened, clicked and typed, as it happens — not a summary afterwards.
Hard limits, set by you
Spending, sending, passwords, deletions, downloads and whole sites can be put off limits. Checked before each step.
Pause and stop, always
Stop ends the run where it stands. Nothing keeps running quietly in the background.
It guards other agents too
Agents and tools you already use can be routed through Sentinel and checked against the same limits.
Connected apps
Works with the tools you already use
Each person connects their own accounts. Everything an agent reads from them is checked for hidden instructions, just like web pages.
- Gmail
- Google Calendar
- Google Drive
- Google Docs
- Google Sheets
- Google Slides
- Outlook
- Teams
- OneDrive
- Word
- Excel
- PowerPoint
- OneNote
- SharePoint
- Notion
- Linear
- Slack
- HubSpot
- Salesforce
- GitHub
- GitLab
- BigQuery
- Snowflake
- Databricks
- Amazon Redshift
- Power BI
Some apps are still being switched on and show as Coming soon inside the app.
Your limits
Everything is allowed except what you switch off
These start switched on. You choose how strict to be, and you can add sites the agent must never touch.
- On by default
Never spend money without asking
Any checkout, payment or subscription stops and waits for you. Set a spend limit and anything above it is refused outright.
- On by default
Never send anything in your name
Emails, messages, form submissions and posts are shown to you in full before they go.
- On by default
Never enter passwords or codes
The agent will not type a password, card number or one-time code. It hands the keyboard back to you.
- On by default
Never delete or cancel
Deleting, unsubscribing, cancelling and archiving are refused.
- On by default
Never download or run files
No files are saved or opened on your machine by the agent.
Said plainly
An agent that reads web pages can be fooled by web pages
Hidden instructions on a page have tricked agentic browsers into buying from fake shops and reading out a one-time code. We will not tell you that is solved.
Sentinel is built for that reality: the limits are checked at the moment the action would happen, so being misled does not turn into money spent, a message sent, or something deleted.
Read the honest limitsReady when you are
Download the desktop app, or open the controls in your browser and watch a run play out against your own limits first.