Intake
Receive community findings using the VDP module Meld.
We build fast, customizable workflows to meet compliance and security requirements. Resolving backlogs in hours rather than days.
Stay compliant
as the rules change.
Select each standard above to see which modules satisfy its requirements.
EU CRA requires 6 of 8 modules.
Scan
Static automated security analysis.
Validate
Agents prove the finding exists.
Triage
Focuses on the important issues.
Human Approval
Stops and waits for a human before proceeding with next steps.
Resolve
Propose a solution to the issue.
Verify
Retest the exploit against new code.
Report
Secure artifacts and documentation.
Mini leads the mighty frontier.
Modern workflows involve security agents finding and resolving issues. Our compact models run fully within your environment, keeping sensitive code under your control.
Agentic vulnerability detection
Benchmark score (%)
Opus 4.6 71.3
Vakttårn 69.2
Minisøk 59.3
GPT–5.5 46.3
Snyk 44.4
VScan Harness, 1000 samples, 2026-06-23
Higher is betterBuilt for agentic security.
Vakttårn matches frontier-model performance on agentic vulnerability detection while substantially outperforming general-purpose SAST engines and openweights models of comparable size.
Score vs. active parameters
Benchmark score (%)
Intelligent + Efficient 
Minisøk : 59.3 on 3 billion active parameters Vakttårn : 69.2 on 6 billion active parameters 
Opus 4.6 : 71.3 on 200 billion active parameters 
GPT–5.5 : 46.3 on 300 billion active parameters
Minisøk : 59.3 on 3 billion active parameters Vakttårn : 69.2 on 6 billion active parameters Opus 4.6 : 71.3 on 200 billion active parameters GPT–5.5 : 46.3 on 300 billion active parametersBillion active parameters (log scale)
Some model sizes are estimates from public information.
Top left is bestA mini-size with mighty performance.
With 50x fewer active parameters, our models run locally on laptops, workstations, and small servers, while your code stays within your network without sacrificing detection performance compared with cloud models.
Automation on your device.
We built our tools to fit in just about every environment, from a single laptop to a cloud deployment.
Connections wherever you work
These tools and workflows are connected through triggers on your device and in the cloud. For example, vscan allows your agent to initiate our code review process using the MCP protocol. If activated, a Vaktex Cloud trigger can execute remote workflows on new VDP reports or examine your code whenever you send a Slack message.
Local and cloud deployments
Don’t want your IP leaving your devices? We let our customers deploy our models locally and efficiently through our inference engine, VSecure. This means your model can run on your laptop or server without any of your code leaving your network.
Tools that find and resolve your reports.
Instead of giving you tickets accross multiple systems, we give you one flow that collects, triages, and resolves your findings.
- Product IDOR via /api/products
- 403 Forbidden bypass on /2fa/enter returns 200 OK
- Socket.IO polling endpoint accepts unauthenticated POST traffic (no 403)
- 403 bypass via 200 on /about
- Authenticated route /address/create accessible without authentication (403 bypass)
- Socket.IO endpoint reflects arbitrary Origin header (CORS misconfiguration)
Reduced resolution backlog
For every finding, we identify its root cause and provide strong integrations between tools, so your reports identify, propose, and implement possible solutions with you, without any of your code leaving your device.
Many reports, one workflow
Maintain your existing workflow regardless of the source of your security reports. We’ve optimized the process to seamlessly collect data from automated tests, VDP, and manual tests, integrating them as if they were a single system, all without the need to switch platforms or tools.
A customized flow, that stays compliant.
Build flows tailored to your team’s needs, while exporting the nessary artifacts to satisfy your auditors and regulators.
Endlessly extensible
Our tools are designed to be modular and extensible, allowing you to create custom security workflows and integrations that connect to our growing ecosystem and fulfill your team’s and auditors’ requirements.
Reporting at all levels
Our services provide unmatched visibility into your security posture through automated reporting and resolution technologies. However, we understand that some reports require a human touch. This is where our industry experts step in to find, validate, and prepare audit-ready reports to meet your industry’s standards.
Product improvement, usage metrics and internet access can each be switched off, in the cloud and on-prem. The switches shown on this plate are an illustration, not a control.
Protect your data, or not,
that choice is always yours.
SOC 2 Type 1
In progress
GDPR
In progress
CASA Tier 2
In progress
SOC 2 Type 2
In progress
In the cloud
Product Improvement
Usage Metrics
And on-prem
Internet
FAQ
What does Vaktex mean?
We selected the name Vaktex as a blend of three concepts reflecting our company's mission and roots. “Vakt” is the Norsk term for a “security guard,” symbolizing both the detection and protection our services provide. “Vak” in Norsk means “awake” in English, describing vigilance against current threats through our ongoing model updates. Lastly, “Tex” is for Texas, where our product was created.
What is vscan?
VScan is our premier security automation tool. It analyzes your code beyond syntax, leverages advanced models to detect vulnerabilities, and provides a secure, isolated environment for agentic resolution.
How do I use vscan?
The easiest way to get started is to read the intro.
Can I fully self-host vscan?
Yes, the vscan client runs locally. If you're interested in self-hosting the models, you'll need an enterprise license and supported hardware.
Do I need external AI subscriptions?
No, our tools rely on proprietary models for detection and resolution, making them incompatible with external providers.
How much does vscan cost?
All Vaktex products are offered as subscriptions due to the continuously evolving nature of cybersecurity threats. For details on our pricing, please visit our pricing page.
What about data and privacy?
Your usage data is used internally when you engage with our promotional plan or opt-in to product improvement. For those in sensitive industries, we offer self-hosting licenses for our models to run on your networks (even air-gapped ones).
Is vscan open source?
No. As developers, we value open-source software and contributions. However, due to the size of AI models, they can't be run locally without powerful hardware. Publishing our weights or training data would primarily advantage competitors rather than ordinary users.
Can I only use vscan in the terminal?
VScan will soon be available as a web app across all platforms.