Security for people and AI agents · Built for MSPs
One platform for MSPs and the businesses they protect: identity, email, laptops, AI tools and AI agents in Microsoft 365 and Google Workspace.
63 countriestried to sign in to the businesses we protect in 30 daysCenteye platform data, Oct 2026
67%of people using AI at work used non-company accountsVerizon 2026 DBIR
Your AI agents
AI agent security Synthwatch · early accessPraxis diagnoses, triages and checks each signal, then lays out the facts in one case. Your team spends its time deciding, not digging through five consoles.
The fixes you allow run at night within your rules. Everything else waits in the morning queue with what happened and what to approve.
Each client gets a plain-English report: what was caught, what was fixed and what is left. Written for you every week.
Already running a sprawl of separate security tools, or an outsourced SOC? See how switching works →
What changed
For twenty years, security protected people and their machines. Today the workforce is people and AI side by side: the AI tools your team connects, the copilots in every app, and the agents that act on their own. Security now has to support both, so the business can keep adopting AI without opening new doors.
Your team pastes, uploads and connects AI tools every day. Company data now leaves through a chat window, not an email attachment.
AI agents sign in, read mail and move files with their own keys. Most have no owner and nobody checking what they do.
A stolen login becomes an AI app on the mailbox, then an agent pulling files. Tools that each watch one piece miss the whole.
The numbers behind the shift
The platform
People security for the humans. Agent security for the AI. And Praxis above both, linking everything into one case and fixing it. Each product works on its own; together they see the whole attack.
Runs on Microsoft 365 · Microsoft Entra ID · Google Workspace · Windows · macOS · Chrome · Edge
Layer 1 · People security
Identity threat detection and response
Stops the attacker who logs in instead of breaking in.
Connects with IronwatchLighthouseLumenSynthwatch Praxis
Explore Centeye GatewatchLayer 1 · People security
Endpoint detection and response
Spots malware and ransomware early, and stops it spreading.
Connects with GatewatchLighthouseLumen Praxis
Explore Centeye IronwatchLayer 1 · People security
Email security
Catches the email your filter lets through.
Connects with GatewatchIronwatchLumen Praxis
Explore Centeye LighthouseLayer 1 · People security
AI governance and registry
Finds the AI tools in use, shows what each can reach, and acts on your rules.
Connects with GatewatchIronwatchLighthouseSynthwatch Praxis
Explore Centeye LumenLayer 2 · Agent security
AI agent security · Early access
Gives every AI agent an identity, an owner and an off switch.
Connects with GatewatchLumen Praxis
Explore Centeye SynthwatchLayer 3 · Above both
Agentic security operations
Diagnoses, triages and checks every case, then gives your team the facts to decide.
Connects with GatewatchIronwatchLighthouseLumenSynthwatch
Explore Centeye PraxisWhat we see
Not a survey. These come from the sign-ins, laptops and accounts Centeye watches for its clients, and each one comes with the fix your team can run.
password guessing, MFA push-bombing and impossible-travel sign-ins at one business, while its laptops raised their own alerts
Tools that watch only sign-ins or only laptops each see half the picture. Centeye watches both for every client, shows them on one page, and puts the fix next to each finding.
tried to sign in to the businesses we protect in 30 days. Only one in five foreign attempts came from a known hosting, VPN or Tor network.
Most attempts came from ordinary home and mobile addresses, the kind a country block or VPN list lets straight through. Centeye judges an address by behaviour, not location: new to this business, trying many accounts, failing in a pattern.
accounts hold an admin role at a 9-person business we watch. The smallest teams have the most admins for their size.
At a small company one stolen admin account is the whole company. Centeye lists every account with an admin role and flags a new admin grant the moment it happens.
Source: Centeye platform data across the businesses Centeye protects, October 2026. Sign-in attempts counted over the last 30 days. No client is named.
Detect · Correlate · Remediate
This is what a modern attack looks like, and why the layers belong on one platform.
Her session is copied from a fake sign-in page and replayed from another country.
The attacker connects an AI note-taker to her mailbox: a way back in that survives a password reset.
The finance agent Dana owns starts reading hundreds of files far outside its normal job.
Kavach links all three. Sessions end, the app is revoked and the agent’s access is cut, each checked at Microsoft. Argus writes it up.
Separate tools would see three small alerts at three different times. Centeye sees one attack on a person, her AI app and her agent, and contains all three in one coordinated response.
You stay in charge
Every fix has a dial, set per action: have Centeye tell you, ask you first, or handle it. Start with everything on “tell me” and turn it up as trust grows.
Centeye is security built for a workforce that is part human and part AI: one platform that knows every person and every agent, sees the attack that moves between them, and helps you shut it down.
