What is Evo Agentic Development Security and why it was created

Snyk — a company used daily by developers at firms like Spotify, Revolut, or Snowflake for code vulnerability checks — officially launched Evo Agentic Development Security (ADS) on June 23, 2026. This is an extension of the Evo by Snyk platform, which the company introduced last autumn as an agentic security orchestration system. While it was more of a vision back then, today it is a finished product with concrete figures, solving a real problem: how to secure software at a time when AI agents, not humans, are largely writing it.

According to data Snyk collected from nearly 10,000 developer environments, 80% of developers run in two or more AI development environments simultaneously, and 50.8% of them have a live connection via MCP servers (Model Context Protocol) — meaning a direct connection of an AI agent to production tools and external systems. In total, Snyk discovered 4,524 unique MCP servers and found that one percent of the busiest environments operate 13 or more MCP servers on a single machine.

Snyk's security team warns: one in twelve developers with an active MCP connection today has a confirmed high or critical vulnerability. And this is no longer just a hypothetical risk.

The PocketOS Case: How an AI Agent Deleted a Database in 10 Seconds

In its announcement, Snyk cites an incident from the PocketOS project, where an AI agent deleted the entire production database, including all backups — in less than ten seconds during a routine bug fix. The agent had access to incorrect credentials, no security guardrails, and no system to stop it. There was no human approval step either.

This event perfectly illustrates why the current security model is no longer sufficient. Traditional application security was built on the assumption that humans write code, changes go through a controlled pipeline, and security scans occur only after the code is complete. In agentic development, this is not the case — risk comes continuously through the tools agents use, the actions they perform, and the code they generate at machine speed.

How Evo ADS Works: Three Pillars of Security

Evo Agentic Development Security does not stand outside the development process like traditional security tools — it is built directly into the workflow of AI-driven development. It operates on three key levels:

1. Agent Supply Chain Security

ADS continuously discovers and maps all MCP servers, skills, tools, and external services that AI agents depend on. Each component is evaluated based on permissions, origin, known vulnerabilities, and corporate security policies. Security teams thus gain, for the first time, a complete overview of everything their AI agents use — and can block unapproved or risky components before they cause harm.

2. Real-time Agent Behavior Management

This is the most significant shift. ADS operates directly within the agent's execution loop — analyzing the session context, user intent, requested actions, and corporate policy before the agent performs an action. It can thus block high-risk operations, restrict permissions, and catch policy violations before they impact systems, data, or infrastructure. Unlike traditional security tools that only report problems after they occur, ADS prevents problems.

3. Code Validation at Creation

Because AI continuously generates code, Snyk has moved security control directly to the moment the code is created — even before it enters a repository or CI pipeline. Checks run asynchronously, and when the code is clean, nothing is fed back into the agent's context window (which saves tokens). When a problem is found, the agent receives only the relevant information needed for correction — without the clutter of old findings.

Evo as a Whole: Not Just ADS

Agentic Development Security is part of the broader Evo by Snyk platform, which covers the entire lifecycle of AI software. In addition to ADS, it also includes:

• AI-SPM (AI Security Posture Management) — inventory and governance of all AI models, agents, workflows, and applications across the organization.
• Continuous Offensive Security — continuous penetration testing of AI systems and red teaming of agents, simulating attacks in the same way as real attackers.

Together, these three components answer the single question every company developing with AI asks today: Can we trust what we deliver to production?

Integration with Existing Tools — Availability for Czech Developers

Snyk Evo ADS is designed to work with the tools developers actually use: Claude Code, Cursor, GitHub Copilot, OpenAI Codex, and others. The security layer integrates directly into IDEs, CI/CD pipelines, and AI coding assistants. The Snyk platform is available worldwide, including a European data center (app.eu.snyk.io), which is crucial for companies that must comply with GDPR and European regulations.

For Czech companies and startups, which are increasingly adopting AI assistants for development (for example, GitHub Copilot has a growing user base in the Czech Republic), Evo ADS offers a way to scale AI development without losing control over security. In addition to enterprise licenses, Snyk also offers a free tier that allows basic code scanning — however, advanced ADS features are part of the enterprise subscription.

What This Means for the Future of Software Development

With its announcement, Snyk opens a new chapter in development security. While AppSec has so far focused on scanning finished code, ADS represents a shift towards securing the entire system that produces the code. It's no longer just about whether the resulting code is secure — it's about whether the tools, data, and processes that lead to it are secure.

The company claims that customers achieve 288% ROI, 80% faster scanning, and 75% faster vulnerability remediation after deploying the Snyk platform. These figures are based on a Forrester Research study that evaluated the impact of the Snyk platform on corporate security.

As AI agents become active creators of software — not just assistants — the question of trust in what they "produce" will become increasingly critical. Snyk Evo ADS is the first tool that attempts to address this systematically — not through post-hoc inspection, but through active real-time management.