We stop AI agents going Rogue
1stProtect stops credential theft, EDR killers and harmful actions by attackers, scripts and AI agents on the device, before they complete, online or offline. Start beside your current EDR, then replace it, and the tools around it, when you’re ready.
Free for 30 days on up to 10 devices. Nothing is blocked until you switch it on. Plans from $30 per device a year.
Online. Every decision is still made on the device.
- PreventedUnknown programReads saved browser passwords
- AllowedAI coding agentEdits files in its own project
- PreventedSigned system toolDownloads and runs a hidden script
- PreventedAI agentReads SSH keys and cloud credentials
- PreventedOffice macroOpens a reverse shell to an unknown host
Partners and customers
“After comparing multiple EDR and endpoint security solutions … we decided to adopt 1stProtect.”
Control your AIs and LLMs.
Coding agents, copilots and local LLMs run with your users’ permissions. 1stProtect decides what each one may do on the device, action by action, whether or not anyone registered it.
Same login. Different actor. Different authority.
| Action | Developer | AI agent |
|---|---|---|
| Read and edit project source | Allowed | Allowed |
| Run tests and builds | Allowed | Allowed |
| Read SSH keys | Allowed | Blocked |
| Read cloud credentials | Allowed | Blocked |
| Run hidden or encoded scripts | Blocked | Blocked |
Illustrative policy. You set the rules for each tool and action.
Any agent, registered or not
Sets what coding agents, copilots and local AI tools may do, including ones nobody registered.
AIControlAvailable now
Keys and credentials stay put
Agents work in the project without reading SSH keys, cloud credentials or saved passwords.
AIControl, CredentialProtectAvailable now
Commands checked before they run
Hidden scripts, reverse shells and risky commands an agent launches are blocked before they run.
ExecutionProtect, ShellProtectAvailable now
Sensitive files stay inside
Controls how sensitive files move between applications, including AI tools.
DataProtectAvailable now
Prompts and MCP tools
Protection against prompt injection, and control over MCP servers and the tools they call.
PromptProtect, MCPProtectComing next
Every case explained
AIProtect explains what an agent tried and why it was stopped, in plain language, when connected.
AIProtectAvailable now
Chosen after hands-on evaluations.
“After comparing multiple EDR and endpoint security solutions and conducting our own evaluation, we recognized the value of 1stProtect’s ‘Prevention by Default’ approach and its protection capabilities. … Based on the results of our evaluation, we decided to adopt 1stProtect.”
“Cybersecurity is entering an autonomous era. Our partnership with 1stProtect brings together advanced cyber defense, AI, and deep enterprise security expertise to help organizations stay ahead of rapidly evolving threats.”
“Our partnership with 1stProtect brings advanced security and control for AI-driven and autonomous environments to organisations across the Philippines. … We are preparing organisations for the autonomous future.”
“We are very pleased to begin offering 1stProtect’s innovative security solutions to our customers. As cyber risks continue to become more sophisticated and diverse, while cloud adoption and digital transformation accelerate, organizations are increasingly seeking security solutions that provide effective protection while remaining practical and easy to operate after deployment. 1stProtect delivers significant value as a next-generation security solution designed to address these evolving needs. Its innovative approach to cybersecurity provides organizations with a powerful way to strengthen their security posture while simplifying the operational challenges associated with endpoint protection. Through our offering of 1stProtect solutions, we will provide comprehensive support throughout the customer journey—from selecting the right solution based on each customer’s specific security challenges, through deployment and implementation, to successful operational adoption. By doing so, we are committed to helping our customers build secure, resilient, and reliable IT environments with confidence.”
Built for teams that hold other people’s secrets.
Built by people who have spent their careers stopping attacks.
Meet the teamKervin Pillay
Chief Executive Officer
25 years in telecoms and enterprise infrastructure, including as an enterprise CTO.
Rafel Ivgi
Co-Founder and Chief Technology Officer
30 years in cybersecurity, from Unit 8200 to hands-on endpoint, data and cloud security engineering.
Headquarters
San Francisco, California
Hosting regions
- United States
- Europe
- Asia
- Middle East
Start beside your EDR. Replace it on your terms.
Land without changing anything. Switch on prevention. Retire your EDR and the tools around it when the evidence is in.
Keeps the keys on the device
of breaches involve stolen or abused credentials somewhere in the attack. 1stProtect stops password dumping, info-stealers and session-token theft before they complete.
Source: Verizon DBIR 2026Stops EDR killers
of about 90 known EDR-killer tools load a vulnerable driver to switch security off. 1stProtect stops the driver from loading.
Source: ESET research, March 2026Judges the action, whoever acts
malicious breaches were AI-enabled in 2026. One policy decides what attackers, scripts and AI agents may do on the device.
Source: IBM Cost of a Data Breach 2026Nothing to crash, nothing to wait for
a device a year, for one agent that replaces several. User mode with no kernel driver, and every decision made on the device, even offline.
See plansOthers observe. We prevent.
Most endpoint tools spot suspicious software and raise an alert. Newer AI-agent tools start by finding and naming every agent. 1stProtect asks a simpler question at the moment it matters: should this action happen at all? It starts beside the tools you run today, and replaces them when you’re ready.
Same attack. Three outcomes.
Detection tools report first and hope someone responds in time. 1stProtect prevents first, then reports.
| Protection | A script arrives in a phishing attachment and runs as the signed-in user. | It opens the browser password store and the saved VPN profile. | It copies the credentials into an archive in a temporary folder. | It uploads the archive to a file-sharing site. | The stolen logins open the VPN from another country. | |
|---|---|---|---|---|---|---|
| Gen 1AntivirusMatches known bad files. Anything new gets through. | Runs | Runs | Runs | Runs | Happens, unnoticed | Breach, and no alert. |
| Gen 2EDRWatches, alerts and responds, after the action has run. | Runs | Runs | Runs | Alert raised | Happens while the alert waits | The alert arrives after the credentials have left. |
| Gen 31stProtectPrevents first, then reports what it stopped. | Runs, nothing harmful yet | Prevented on the device | Reported, with the full story | Never happens | Never happens | Prevented first, then reported. Nothing to clean up. |
Step 1: Script starts A script arrives in a phishing attachment and runs as the signed-in user.
- Attack step succeeds
- Alert raised
- Prevented on the device
- Reported
- Never happens
Illustrative example. Compare control models
Resolved on the device, long before it reaches the SOC.
Detection-first tools collect telemetry, raise an alert and wait for an analyst, anywhere from minutes to days later. 1stProtect resolves the threat inline, on the device, at the moment it happens. Its AI investigator then explains what was stopped, so your team reviews a closed case instead of chasing an open one.
- ActionAn attacker, script or agent acts
- TelemetryEvents are sent for analysis
- DetectionRules and models flag it
- Alert queueIt waits for an analyst
- ResponseTriage, contain, reset
Watch it stop a credential theft.
A script on a protected device tries to copy five kinds of saved credentials. Every copy is blocked before it completes.
- Watch every risky actionReading passwords and keys, starting programs, running scripts, opening connections.
- Decide on the deviceEach action and its target are judged locally by one on-host Sigma engine, with no trip to the cloud.
- Stop the harmful onesHarmful actions are blocked before they complete. Everything else runs as normal, and the console records why.
What happens in the video
The script goes after a password vault, browser cookies, FTP and VPN profiles and an SSH key. Each copy is blocked before it completes and the user sees a short notice. The console records each attempt with the full chain from device to file.
Real console, demo environment.
A dedicated engine for every step an attack needs.
One lightweight agent. Each engine stops one kind of harmful action.
Dedicated prevention engines
- CredentialProtect
- DriverProtect
- ExecutionProtect
- ShellProtect
- AIControl
- RansomProtectRoadmap
- WiperProtectRoadmap
Stealing credentials
Password dumping, info-stealers, session-token theft and exposed keys.
CredentialProtect
Misusing trusted tools
Built-in system tools and signed apps running hidden scripts, payloads and shells.
ExecutionProtect, PayloadProtect, ShellProtect
Running untrusted software
Unapproved apps, known malware and vulnerable drivers.
AppControl, MalwareProtect, DriverProtect
AI agents overstepping
Local agents and copilots taking actions nobody approved, registered or not.
AIControl
Moving data out
Sensitive files moving between apps, and screen leaks you can trace.
DataProtect, ScreenTrace
Moving sideways
Stolen identities reused across machines, and remote sessions nobody approved.
LateralProtect
One agent that keeps growing.
Every release adds coverage to the same agent, with no new install. Teams use it to retire separate tools one at a time, so they manage fewer vendors, pay for fewer licences and run fewer agents on every device.
Fewer vendors to manage
One contract, one console and one support line instead of several.
Lower total cost
Retire overlapping licences as coverage arrives, instead of buying another product for each new need.
One footprint
One install, one update cycle and one policy model on every device.
Lighter devices
Fewer agents competing for CPU, memory and disk on each machine.
1stProtect covers 3 of the 4 tools you picked, today.
That’s up to 3 agents you could retire from every device, along with the contracts behind them.
1 more is on the roadmap, in the same agent.
Your EDR: start beside it, then retire it in stages once 1stProtect covers what you rely on.
Coverage means 1stProtect has an engine for that need. Compare features with your requirements before you retire a tool.
Starts beside your EDR. Replaces it when you’re ready.
Install beside the tools you run today, with no kernel driver to collide. Prove it on your own devices, then retire your EDR and the point tools around it, one at a time.
- AuditInstall on up to 10 devices in audit mode and get a report of everything it would have blocked.
- EnforceMove engines from Detect to Prevent, policy by policy.
- ConsolidateRetire the tools 1stProtect makes redundant. One agent replaces several.
Does 1stProtect replace my EDR?
Yes, in stages. It starts beside your current EDR and adds prevention at the moment of action. Once it’s proven on your devices, you retire your EDR and the overlapping tools.
Does it work offline?
Yes. Every prevention decision is made on the device, so protection continues on disconnected and air-gapped machines. The AI investigator needs a connection today.
How much does it cost?
Plans start at $30 per device per year. The price depends on the bundle, device type, number of devices and term. See plans.
See what it would have stopped.
Run 1stProtect free for 30 days on up to 10 of your own devices. You get a report of everything it would have blocked, and nothing changes until you decide.
- Audit mode first
- Runs beside your current EDR
- A report of every would-be block