AI Agent Security
Runtime security for AI applications and agents, formerly Lakera Guard: inventories agents across cloud and low-code platforms, scores each agent's risk, restricts which tools and MCP servers it may call, and screens prompts, responses and tool use for prompt injection, data leakage and unauthorised actions.
Find alternatives to AI Agent Security
What it does
- Guardrails and safety
- Data security
- Threat detection and response
Features 6
- Agent Discovery and InventoryIdentifies agents running across supported cloud and low-code platforms.
- Per-Agent Risk AssessmentAssesses risk across agent metadata, usage, tools, skills, MCP servers and connected components.
- Tool and MCP Access ControlAllow and deny controls for which tools and MCP servers agents can use at runtime.
- Agent Action ControlEvaluates agent tool calls and actions in context to block unsafe or unauthorized behavior.
- Prompt Attack and Data ProtectionDetects prompt attacks, indirect injection and sensitive data exposure in prompts, responses and tool use.
- Unauthorized Action DetectionDetects agent actions outside user intent, policy or approved workflows.
Sources
- Pricing
-
Community: $0 per month, 10k requests a month, 8k-token prompts, SaaS, EU residency. Enterprise: "Let's chat!", flexible requests, SaaS or self-hosted, SSO, RBAC, SIEM integration, EU/US residency.
- Description
-
Official documentation · 10 Oct 2026
Features: "Agent Discovery and Inventory", "Per-Agent Risk Assessment", "Tool and MCP Access Control", "Agent Action Control", "Prompt Attack and Data Protection", "Unauthorized Action Detection".
- What it does
-
Official documentation · 10 Oct 2026
"Help detect prompt attacks, indirect injection, and sensitive data exposure across agent prompts, responses, and tool use."
- Sold within
-
Self-serve: "The default pricing plan (Community) provides free access to a limited number of requests. Contact sales to update your plan and get access to advanced features." The Lakera Guard platform is the product that lakera.ai/lakera-guard now redirects to as AI Agent Security.
- Name
-
Official documentation · 10 Oct 2026
A-Z index and page breadcrumb "AI Agent Security".
- URL
-
Official documentation · 10 Oct 2026
Check Point's own product page, in the products A-Z index under AI Security. lakera.ai/lakera-guard now redirects to lakera.ai/ai-agent-security, which carries the same title, "Secure AI Agents from Discovery to Runtime".
- Deployment
-
Hosting: Community "SaaS"; Enterprise "SaaS or Self-hosted".
- Interfaces
-
"API Support: Included in Community"; lakera.ai lists "API-first architecture".
- Intended for
-
Official documentation · 10 Oct 2026
"trusted by industry leaders, from Fortune 500 companies to startups"; developers integrate it through the API.
Also from Check Point Software Technologies
-
ThreatCloud AI Platform
Check Point's threat-intelligence and prevention engine, fed by telemetry from 150,000 connected networks and millions of endpoints, that runs more than 50 AI and traditional engines for zero-day phishing, unknown malware, anomaly and DNS-tunnelling detection and shares verdicts across Check Point's network, cloud, endpoint and workspace products.
-
Workforce AI Security Application
Discovers which AI apps, browser extensions, coding assistants, desktop agents and MCP tools employees use, sanctioned or shadow, and applies per-app policy and AI-powered data-loss prevention that redacts sensitive data from prompts, uploads and files, deployed through a browser extension or desktop agent.
-
AI Red Teaming Developer tool
Automated adversarial testing of models, AI applications and agent endpoints that runs broad scans and targeted campaigns across 400+ foundation models, generates context-specific attack inputs, and re-runs them after model or prompt changes to catch regressions in safety and security.
-
Check Point AI Assist Application
Generative AI assistant for security administrators, formerly AI Copilot, that knows a customer's policies, access rules, objects and logs and answers in chat, creating and updating firewall and DLP rules, guiding incident investigation and writing playbooks across the Check Point platform.
-
Multi-agent system for network security operations in which specialist AI agents, coordinated by a master orchestrator over a live knowledge graph of the customer's topology and rulebase, turn plain-language intent into firewall rules, tighten over-permissive policy, troubleshoot failing flows and map changes to DORA, PCI-DSS and NIST, with high-impact changes routed for human approval.
-
Agentic Exposure Validation AI agent
Uses AI agents that reason like an attacker to test whether each vulnerability, leaked credential or exposed asset is actually exploitable, enriching findings with live threat intelligence, checking existing controls, running safe attacker-style validations reviewed by a separate AI safety reviewer, and returning only proven exposures.
-
AI Factory Firewall Infrastructure service
Check Point security software for AI data centres and neocloud providers that runs on NVIDIA BlueField DPUs beside GPU servers, providing network, host and code-level protection against model theft, data poisoning, model inversion and exfiltration without taking compute from AI workloads, with multi-tenant management.
-
AI Network Firewall Application
AI-traffic inspection delivered in Check Point firewall software release R82.20, so an existing physical or virtual Check Point firewall discovers employees' AI apps and agents, governs MCP traffic, stops sensitive data leaving in prompts and blocks prompt injection inline before it reaches an LLM.
-
Check Point Email Security Application
Check Point Email Security uses AI to investigate and prevent email threats.
-
Check Point XDR Platform
Check Point XDR correlates security events to uncover and prevent cyberattacks using AI.
Open Check Point Software Technologies in the directory
Something wrong here? Send a correction — quote this product id: check-point-ai-agent-security.