AgentCloak: advanced data leak prevention for MCP and A2A AI agents
Comply with AI data protection and cross-border Sovereign AI requirements
Bidirectional data cloaking and uncloaking
Tuned AI detects and generalizes, tokenizes, hashes, and masks protected data, and then uncloaks data in responses
Governed digital twins secured by identity
Maintain secure digital twins with custom schemas governed by identity context from identity servers and authorization tokens.
Seamless integration with MCP and A2A
Immediately protects agents and autonomous agents as an MCP Server, MCP proxy, A2A proxy, REST API or web services proxy.
Seamlessly substitute sensitive data and restore original values based on identity context
- Realtime tuned AI detects PII personal data, PHI health data, and PCI DSS account numbers.
- Comply with EU AI Act data minimization requirements.
- Replace protected data with flexible tokenization, hashing, and masking strategies.
- Restore values only with proper identity context.
- Works with both structured and unstructured MCP and A2A requests and responses.
Protect with next generation, AI-powered data generalization
- Fast and tuned AI generalizes personal and health data.
- Identifies multiple participants in a single conversation, such as “I’m writing about my father”.
- Generalizes ages and weights into age and weight brackets.
- Generalizes health terms like “headache” into generic terms like “minor neurological condition”.
- Generalize addresses into cities, states and provinces, countries, or regions.
Maintain a secure digital twin and govern with identity context
- Secure digital twin maintains protected data profiles.
- Memory of context history for contextual processing.
- Utilizes identity context from existing identity servers to manage cloaking and uncloaking policies.
Data protection trusted by top multinationals and approved by the most rigorous regulators
Ensure Sovereign AI for cross-border agents
- Invoke global AI agents without transmitting protected data across borders.
- Uses tuned Sovereign AI to detect protected data with AI deeply trained in local language and vernacular.
- Seamlessly add data residency to MCP and A2A agentic flows.
- Translate data with Sovereign AI from local language to global language.
- Sovereign digital twins that are fully managed in Sovereign clouds or on-premises.
- Available in China, Saudi Arabia, UAE, Germany, India, and other jurisdictions with AI and data sovereignty regulations.
Invoke protection for clients and servers, human in the middle and autonomous agents
- Immediate protection secures agentic flows while ensuring AI remains fast and reliable.
- Data Leak Protection protects unnecessary sensitive data from reaching AI servers.
- Data Loss Prevention protects data from escaping AI servers.
- AgentCloak can protect both clients and servers concurrently.
- Works with human in the middle agents and fully autonomous agents.
Seamlessly integrates into agentic flows, training, fine-tuning, and RAG
- Manage policies from a centralized dashboard and update in realtime
- Customize security rules and policies for different identity roles, apps, and use cases
- Invoke as an MCP Server from any agent.
- Integrates as a proxy for MCP, A2A, or web services.
- Train agents with cloaked digital twins and uncloak data with identity-based authorization
- Detailed logging and OpenTelemetry support for full observability
Deploys as a secure edge service
Multi-tenant
Single tenant in any cloud
AWS Outposts
On-premises
Active-active failover available for all deployment models
Download the AgentCloak technical white paper
Explore the technical details of how AgentCloak safeguards your agentic flows