Access is abused. Alerts arrive late. Intent remains a blind spot.
.png)
Capsule One:
the System One Model for Agentic Intent Security
Built to decide. Not to generate another answer.
Capsule One is Capsule’s own fine-tuned small language model for runtime security decisions. Instead of generating a long response, it scores whether a proposed tool call fits the authorized task, fast enough to decide at the tool-call boundary.
Together, we fine-tuned NVIDIA Nemotron detectors and tested them on the public StepShield benchmark.
Read the researchSecure intent. Not just access.
Access answers “Can this agent use the tool?” Intent security asks “Does this action belong to the task?”
Evaluate agent behavior in context: its identity, instructions, session history, and the action it is about to take.
Detect when a proposed tool call crosses the authorized boundary, then allow, flag, or block it according to policy.
Connect runtime protection to your AI workflows through supported integrations, without redesigning your architecture. Monitoring and blocking capabilities vary by integration.
.png)
.png)
.png)
.avif)
.avif)
.avif)
.avif)
.avif)
From intent to action. With a security decision in between.
Follow a proposed tool call from its authorized task to Capsule’s runtime decision.
Secure runtime for any AI Agent in your workflow
AI Agent Builder Platforms
AWS Bedrock
Azure Foundry
GCP Vertex
AI Coding Agents
Claude Code
Cursor
Github Copilot
Enterprise AI Agents
ChatGPT Enterprise
Microsoft Copilot Studio
Salesforce Agentforce
See every agent. Secure every action.
Guardian Agent
Capsule acts as the enterprise’s intelligent, always-on guardian agent, continuously discovering, observing, and securing AI agents across the organization while proactively detecting threats, vulnerabilities, and suspicious behavior in real time.
Agent Security Graph
The Capsule Agent Security Graph maps how agents think, act, and interact at runtime by analyzing relationships between agents, tools, data, and actions—revealing risky paths, control gaps, and emerging threats in a clear, intuitive view.
Deep Observability
Gain deep, real-time visibility into agent behavior, including actions, decisions, and execution paths—providing continuous insight into how agents operate in production and enabling faster investigation, governance, and safe scaling.
Runtime Protection
Enforce security and governance policies in real time, before actions are executed. Capsule detects and blocks unsafe, unintended, or risky agent behavior in real time, preventing incidents without slowing agents or engineering teams.
Agentic Policy Control
Continuously understand who the agent is, who owns it, who can access it, and what it’s allowed to do, maintaining clear ownership, least privilege, and accountability across all agent environments.
Capsule acts as the enterprise’s intelligent, always-on guardian agent, continuously discovering, observing, and securing AI agents across the organization while proactively detecting threats, vulnerabilities, and suspicious behavior in real time.
The Capsule Agent Security Graph maps how agents think, act, and interact at runtime by analyzing relationships between agents, tools, data, and actions—revealing risky paths, control gaps, and emerging threats in a clear, intuitive view.
Gain deep, real-time visibility into agent behavior, including actions, decisions, and execution paths—providing continuous insight into how agents operate in production and enabling faster investigation, governance, and safe scaling.
Enforce security and governance policies in real time, before actions are executed. Capsule detects and blocks unsafe, unintended, or risky agent behavior in real time, preventing incidents without slowing agents or engineering teams.
Continuously understand who the agent is, who owns it, who can access it, and what it’s allowed to do, maintaining clear ownership, least privilege, and accountability across all agent environments.
