Straight answers.

Security for a workforce that is part human and part AI. Centeye protects your people (sign-ins, laptops, email and how they use AI), governs your AI agents, and runs the response with Praxis, across Microsoft 365 and Google Workspace.

People security: Gatewatch (identity), Ironwatch (devices), Lighthouse (email) and Lumen (AI governance and registry). Agent security: Synthwatch. Above both: Praxis, the agentic security operations layer, with its two AI agents, Kavach and Argus.

Small and mid-sized businesses that run IT in house, and the MSPs who run IT for them. Same product, same console.

Yes. Every fix has a dial, set per action: tell me, ask me first, or handle it. Risky steps never run on their own, and every step is checked at Microsoft or Google afterwards and recorded.

Microsoft 365 and Google Workspace connect by API, with nothing to install. Laptops get a light agent for Windows and macOS. Lumen adds a Chrome and Edge extension and a Windows sensor for AI use on the device.

Microsoft 365 and Entra ID, Google Workspace, Copilot and Gemini, Windows and macOS, Chrome and Edge; standard PSA tools for tickets and RMM tools for devices; Slack, Microsoft Teams and email for alerts and approvals.

Centeye is sold mainly through MSPs, in packages that match what each client needs, with Praxis included. Companies running their own IT can start a free trial or talk to us. See the pricing page for how it works.

Yes. That is who it is built for. Praxis does the watching and the first response, so one person can cover nights and weekends without a security team.

Good, keep them. Ironwatch uses Defender’s detections, and Gatewatch adds the response Microsoft leaves manual, like removing an attacker’s sign-in method. It also covers Google Workspace.

Yes. The same product runs through MSPs. Send them to centeye.io/for-msps, or ask us to talk to them.

No. Praxis does the first-line work: triage, linking, a proposed fix and a check afterwards. You set the dial per client and per action and take the approvals. If you have analysts, approvals go to them.

Whatever your dial says. Let the clear-cut fixes run on their own and nights are covered. Everything else waits with its evidence.

It can, or it can run beside them while you compare. Ironwatch works with Microsoft Defender rather than fighting it, and Gatewatch works alongside Entra ID P2.

Route their approvals to them, or keep them for yourself. The dial is set per client and per action, and the client’s IT person sees the same evidence you do.

Per client, in packages that match what each client needs: secure the people, get AI under control, secure the AI workforce, or the whole platform. Praxis is included. Talk to us and we will shape it around your client list.

P2 is a good sensor, but only on Microsoft and only where you pay for it, and steps like removing an attacker’s sign-in method are still manual. Gatewatch detects and responds on Microsoft and Google and works alongside P2.

From signals that line up: an odd sign-in, then a new inbox rule, a new app grant, or mail forwarded outside. Gatewatch joins them into one case instead of four small alerts.

Nothing, unless their account is locked or they are asked to confirm a sign-in was them.

No. Ironwatch picks up Defender’s detections and adds the response Defender doesn’t run on its own, like cutting a machine off the network, restoring changed files and linking the event to the login that caused it.

Light. It rolls out through your device-management tool and sends events, not files, unless an investigation needs a sample.

Windows and macOS laptops, desktops and servers.

It is detection and response that runs around the clock inside rules you set, with Praxis doing the first-line work. Your team, or your MSP, makes the calls the rules leave to people.

No. Lighthouse connects by API and works after delivery, behind the filter you already run.

Mostly messages with no bad link or file: a fake supplier asking to change bank details, a lookalike domain, a display name copied from your CEO.

It reads a message to judge it and does not keep the body after the verdict.

Yes. One step pulls a whole campaign from every inbox it reached.

No. It is about knowing which AI tools touch company data and deciding which to keep. Most companies approve a short list, coach people toward it, and block only what is risky.

No. The browser extension and the device sensor never read files, messages or the words typed. They see which AI service, which person and how much data.

Observed transfer: data was sent. Possible disclosure: sent to an unapproved app or a personal account. Confirmed exfiltration: only when a named person reviews it and records it, never the software on its own.

No. Lumen runs on Microsoft 365 Business Premium and Google Workspace Business plans.

Lumen covers how people use AI: the browser, the device and the data going out. Synthwatch covers the AI agents themselves: their identity, their access, their behaviour and their posture. They share one register and one console.

No. Synthwatch runs on Microsoft 365 Business Premium and Google Workspace Business plans.

No. It stays outside the agent and acts through identity, OAuth grants and access.

Synthwatch is in early access. We are onboarding a small number of businesses and MSPs now; start a free trial or book a call to join.

No. It is the part of Centeye that links signals into one case, proposes the fix and runs it inside your rules. You can ask it questions, but its job is to work the case.

The two Praxis agents. Kavach is the analyst on shift: it watches, links signals into one case and fixes it within your rules. Argus is the advisor: it explains each case in plain English and writes the weekly report.

What your dial says. Fixes you allowed run and are checked; everything else waits with its evidence for the first person to look, or for one tap on a phone.

No. Risky steps never run on their own, and every step is recorded with who, what, why and when.