Public-source research
Vendor evidence
Review what vendors say publicly, the exact quoted source text, the security requirement each statement may support, and what still needs verification.
Research library coverageCounts describe the research workflow, not vendor quality or product effectiveness.View details
Evidence records1280
Source checked863
Needs verification0
Source captured0
No supporting claim found411
Excluded from evidence6
Wiz claims runtime detection and response for prompt injection, rogue agents, malicious AI behavior, and Model Context Protocol (MCP)-connected AI systems using sensor and cloud telemetry.
Stop AI-native threats including prompt injection, rogue agents, and malicious AI behavior at runtime.
Related framework references (5)
Microsoft Purview claims data loss prevention (DLP), sensitivity-label, risky-interaction, and unethical-behavior policies across prompts and responses for supported AI applications and agents.
This recommendation creates a policy to help calculate user risk by detecting risky prompts and responses.
Related framework references (5)
Noma claims identity-aware runtime enforcement across Model Context Protocol (MCP) connections, prompts, tool calls, data access, agent actions, model responses, and application programming interface (API) traffic.
Every MCP connection is checked against the registry the moment it’s established.
Related framework references (5)
WitnessAI claims bidirectional runtime defense and organization-wide allow or block policy enforcement across prompts, responses, agent actions, tools, Model Context Protocol (MCP) servers, models, integrated development environments (IDEs), and applications.
A security team approves which MCP servers and tools agents may use, and enforces that policy organization-wide.
Related framework references (5)
Netskope claims an AI Gateway and Agentic Broker with real-time decoding, inspection, data loss prevention (DLP), guardrails, and granular blocking policies for prompts, responses, Model Context Protocol (MCP) servers, tools, and traffic.
Advanced real-time protection policy controls.
Related framework references (5)
Zscaler claims Model Context Protocol (MCP) and agent-to-agent (A2A) brokers, fine-grained agent access policy, prompt hardening, AI guardrails, and runtime protection across enterprise AI agents.
Secure agentic communications through MCP and A2A brokers and enforce fine-grained access policies across every enterprise AI agent.
Related framework references (5)
Check Point AI Agent Security claims real-time Guard application programming interface (API) screening and enforcement across prompts, tool calls, tool responses, tool descriptions, data leakage, content violations, and off-policy agent behavior.
Real-time screening and flagging of prompt attacks, data leakage, content violations, and off-policy agent behavior through the Guard API.
Related framework references (5)
Lasso claims an Model Context Protocol (MCP) security gateway and intent-aware runtime policy enforcement across agent actions, tools, application programming interfaces (APIs), external connections, indirect prompt injection, memory poisoning, permissions, and data loss prevention (DLP).
Lasso's open-source MCP Gateway provides a security layer for MCP connections.
Related framework references (5)
Pangea AI Guard claims application programming interface (API) and gateway-integrated enforcement across prompts, model responses, retrieval-augmented generation (RAG) ingestion, agent plans, tool inputs, and tool outputs using configurable block, report, redact, encrypt, and defang actions.
Recipes can be configured to block, report, redact, encrypt, or defang sensitive or malicious content.
Related framework references (5)
F5 AI Guardrails claims model-agnostic runtime security for models and agents with real-time input and output protection, data-loss prevention, adversarial-threat blocking, customizable policy, and observability.
Comprehensive runtime security for AI models and agents.
Related framework references (5)
Oasis claims intent-aware runtime policy that converts prompts, tool calls, and action plans into short-lived least-privilege sessions with allow, warn, deny, or step-up enforcement before execution.
Every interaction is turned into a short-lived, least-privilege session with full accountability.
Related framework references (5)
Astrix claims policy-driven runtime access enforcement for agents using precisely scoped, short-lived, just-in-time credentials and pre-approved connectivity.
Every agent and workload gets policy-driven, short-lived credentials delivered with precisely scoped and just-in-time access.
Related framework references (5)
Entro claims real-time identity policy across agents and NHIs, just-in-time scoped access, intent monitoring, anomaly detection, and Model Context Protocol (MCP) session auditing for prompts, servers, and agent contacts.
Entro’s AI Detection and Response monitors agent intent in real time, and catches threats at the identity layer.
Related framework references (5)
Aembit claims an Model Context Protocol (MCP) Identity Gateway that validates workload identity, enforces per-request policy, approves or denies access, exchanges credentials, and logs agent-to-resource communications.
The gateway authenticates the agent, enforces policy, and performs token exchange.
Related framework references (5)
Prompt Security claims an AI and Model Context Protocol (MCP) Gateway that inspects requests, responses, prompts, templates, agent actions, and server interactions in real time with allow or block policy, threat intelligence, data loss prevention (DLP), and endpoint enforcement.
Inspecting every request and response in real time to protect sensitive data and information.
Related framework references (5)
Harmonic claims a locally installed Model Context Protocol (MCP) Gateway and endpoint controls that intercept Model Context Protocol (MCP) traffic, inspect prompt and tool interactions, block risky actions, prevent sensitive-data exfiltration, and apply intent-aware policy in under 200 milliseconds.
Transparently intercepts all MCP traffic enabling security teams to discover what clients and servers are in use, enforce granular policies to block risky actions.
Related framework references (5)
Cyberhaven claims endpoint-resident runtime observability and policy enforcement across agent files, application programming interfaces (APIs), Model Context Protocol (MCP) servers, tools, outputs, and sensitive-data movement using behavioral context and data lineage.
The third pillar is runtime policy enforcement.
Related framework references (5)
Nightfall claims protocol-level interception, full request visibility, granular server and tool control, sensitive-data detection, auto-redaction, blocking, anomaly detection, and quarantine across Model Context Protocol (MCP) prompts, files, application programming interface (API) calls, responses, and tools.
Monitor every MCP tool call in real-time.
Related framework references (5)
Island claims one policy engine across AI browser, desktop, extensions, network, prompts, outputs, agents, and more than 500 governed Model Context Protocol (MCP) integrations with prompt-injection mitigation and human checkpoints.
MCP Gateway with governed access to 500+ integrations.
Related framework references (5)
LayerX claims real-time last-mile monitoring, detection, blocking, data loss prevention (DLP), and adaptive policy across prompts, agent actions, data exchanges, AI applications, browsers, desktops, integrated development environments (IDEs), extensions, and on-device agents.
Monitor, detect, block, and govern AI interactions in real time with adaptive controls.
Related framework references (5)