Snyk launches cloud remediation agent with high merge success rates

Snyk has released a cloud remediation agent designed to automate the fixing of security vulnerabilities across enterprise codebases. In internal testing on a subset of more than 100 repositories, the agent achieved a 93% merge success rate. Nair noted that using Snyk's harness alongside Cursor resulted in a 90% merge success rate, compared to 60% with Cursor alone, and was 61% more token efficient. The system utilizes a benchmark called VULNBump to evaluate performance and incorporates global context from millions of repositories to determine the safest dependency upgrade paths.

Snyk has released skills for Cursor and other coding agents, with customers like Labelbox and Relay Networks reporting they reached "backlog zero" using these tools. Nair stated that 1,500-plus enterprise customers have adopted Snyk Studio, which feeds security context directly into the agentic loop. The company is also using frontier intelligence for triage to reduce false positives, allowing the remediation engine to focus on exploitable issues rather than prioritizing all findings. Nair presents the remediation agent as the most promising application of agentic AI for addressing the growing security backlog, arguing that purpose-built harnesses significantly outperform general-purpose coding agents in security-specific tasks.