Dreadnode Launches 2.0, the First Complete Infrastructure Platform for Security Agents
By Dreadnode Team
The platform equips security teams with a suite of tools to build, evaluate, and deploy security agents with confidence
Tuesday, March 24, 2026 – Dreadnode, the infrastructure platform for the security stack, today announced general availability of Dreadnode 2.0. The release introduces hosted evaluations, a model training and optimization pipeline, advanced AI red teaming, capability registries, managed sandboxes, and a new visual frontend — everything required to close the loop from proof-of-concept to production agents.
Dreadnode was founded with the mission to prove AI could perform offensive security operations. That thesis has been validated across use cases, and now the opportunity is bigger than offense versus defense. AI is going to change the entire security stack. Dreadnode will continue to prioritize offensive security and AI red teaming, but the 2.0 release signals what’s required to make AI-native security a reality.
“Models have proven they can perform at par, or exceeding, human capability. But capability without infrastructure won’t solve security,” said Will Pearce, CEO and Co-Founder of Dreadnode. “AI transformed software engineering and spawned an ecosystem of infrastructure companies to accelerate the domain. Now, security is undergoing the same transformation, but purpose-built infrastructure didn’t exist. So, we built it.”
“Every team we talk to is working on accelerating with this tech. The question isn’t if agents are the future—it’s whether they have the tools to build, deploy, and improve them at scale,” said Nick Landers, CTO and Co-Founder of Dreadnode. “We’re the infrastructure layer that makes that possible, with control of your data, and tooling that matters.”
What’s New in 2.0
- Resource Hub: Tap into a repository of approximately 1,600 security tasks, alongside custom datasets, models, and packaged offensive capabilities of agents, skills, and tools. Publish and version your own work privately, or share them with the community.
- Advanced AI Red Teaming: An algorithmic approach to AI security probing, safety assessments, agentic security testing, multi-modal and multi-lingual probing, and model evaluations.
- Hosted Evaluations: Bring your own environments, scorers, and metrics for measuring agent performance without the operational burden of rate limits, timeouts, and monitoring.
- Training & Optimization: Modules for model fine-tuning and prompt optimization, enabling teams to train on their own operational data with on-premises inference.
- Worlds: Synthetic network environment generation that simulates protocols and produces trajectories for agent training and testing.
- Managed Sandboxes: Safe, segmented, and scalable compute environments for agent execution with remote compute and guardrails built-in. Scale agentic capabilities, no matter the use case.
- Updated Frontend: A new visual interface for Dreadnode’s agent framework, allowing teams to run capabilities, like security operations, network ops, web app testing, reversing, and more. Deploy locally or in sandboxes alongside the TUI.
- Tracing & Observability: Deep metrics on agent behavior, going beyond session logs to enable data-driven iteration.
Dreadnode’s approach is deliberately different from both closed security agent vendors that deliver single-purpose agents and generic agent frameworks that weren’t built for security’s unique requirements. Rather than building point solutions (e.g. AI for web app testing, AI for network operations) the platform provides infrastructure that supports any security agent use case—offense and defense. Teams bring their domain expertise; Dreadnode provides the platform to scale it.
The platform is generally available at dreadnode.io with two core plans:
- Pro: For builders, practitioners, and teams. Full-featured SaaS with a pay-as-you-go and only for what you need pricing model. Start for free.
- Enterprise: For organizations who require scaled agentic workloads and custom capabilities. On-prem deployment and dedicated support from the Dreadnode team.
About Dreadnode
Dreadnode is the infrastructure platform for building, evaluating, and deploying security agents with confidence. Built by offensive security and AI red team operators from NVIDIA, Microsoft, NetSPI, Meta, and Cohere, Dreadnode provides the complete toolkit teams need to operationalize AI at every step of the development and deployment process. Dreadnode envisions a world where agents are the trusted standard for security operations. A world where security costs less and does more.
Learn more at dreadnode.io.
Contact
Tori Norris
Head of Marketing, Dreadnode
Signal: Tori.777