# Agent Security Index

> An independent, vendor-neutral directory of 78 AI agent security tools in 6 categories. Every fact comes from the vendor's own site, GitHub or a linked public source; blank fields mean not verified. Listings last verified 2026-10-17.

No vendor pays for ranking. Tools are listed alphabetically within each category. Paid placements, when they exist, are labeled. No lab test results have been published yet.

## Data

- [tools.json](https://agentsecurityindex.com/data/tools.json): every listing as JSON (name, url, category, description, model, license, pricing, vendor, parent_or_acquisition with parent_source, github_repo, last_verified)
- [llms-full.txt](https://agentsecurityindex.com/llms-full.txt): every listing in full as plain text
- [Sitemap](https://agentsecurityindex.com/sitemap.xml)

## Categories

- [Prompt-injection firewalls & guardrails](https://agentsecurityindex.com/category/firewalls-guardrails/): Filters that sit in front of or around an LLM and block prompt injection, jailbreaks, data leaks and unsafe output. (16 tools)
- [MCP gateways & scanners](https://agentsecurityindex.com/category/mcp-security/): Gateways that control which MCP tools an agent can reach, and scanners that check MCP servers and agent skills for tool poisoning and other flaws. (13 tools)
- [Agent runtime security & posture](https://agentsecurityindex.com/category/runtime-posture/): Platforms that discover AI agents across a company, watch what they do at runtime and enforce policy. (11 tools)
- [AI red teaming & evals](https://agentsecurityindex.com/category/redteam-eval/): Tools and services that attack models and agents on purpose to find weaknesses before attackers do. (14 tools)
- [Agent identity & authorization](https://agentsecurityindex.com/category/identity-authz/): Identity, credential and permission systems that give agents their own scoped, auditable access. (13 tools)
- [Agent audit & observability](https://agentsecurityindex.com/category/audit-observability/): Tracing and logging platforms that record what agents did, step by step, for debugging and audit. (11 tools)

## Prompt-injection firewalls & guardrails

- [Akamai Firewall for AI](https://agentsecurityindex.com/tools/akamai-firewall-for-ai/): Edge or API-deployed LLM firewall that blocks prompt injection, jailbreaks, toxic output and data exfiltration. (Commercial, Akamai)
- [Amazon Bedrock Guardrails](https://agentsecurityindex.com/tools/amazon-bedrock-guardrails/): Configurable safeguards for Bedrock and other models: content filters, denied topics, PII redaction and prompt-attack detection. (Cloud service, Amazon Web Services)
- [Azure AI Content Safety Prompt Shields](https://agentsecurityindex.com/tools/azure-ai-content-safety-prompt-shields/): Azure service that detects user prompt attacks and indirect (document) prompt injection. (Cloud service, Microsoft)
- [Cloudflare AI Security for Apps (Firewall for AI)](https://agentsecurityindex.com/tools/cloudflare-ai-security-for-apps/): WAF detection that flags prompt injection, PII and unsafe topics in AI application traffic at Cloudflare's edge. (Cloud service, Cloudflare)
- [F5 AI Guardrails (formerly CalypsoAI)](https://agentsecurityindex.com/tools/f5-ai-guardrails/): Runtime guardrails for generative AI apps; calypsoai.com now redirects here. (Commercial, F5)
- [Google Cloud Model Armor](https://agentsecurityindex.com/tools/google-cloud-model-armor/): Google Cloud service that screens prompts and responses for prompt injection, sensitive data leaks and harmful content. (Cloud service, Google Cloud)
- [Guardrails AI](https://agentsecurityindex.com/tools/guardrails-ai/): Open-source framework and validator hub for checking and correcting LLM inputs and outputs. (Open source + commercial, Guardrails AI)
- [Lakera Guard](https://agentsecurityindex.com/tools/lakera-guard/): Real-time API that screens LLM inputs and outputs for prompt injection, data leakage and policy violations. (Commercial, Lakera)
- [Lasso Security](https://agentsecurityindex.com/tools/lasso-security/): AI security platform for visibility, control and protection across AI models, agents and apps. (Commercial, Lasso Security)
- [LLM Guard](https://agentsecurityindex.com/tools/llm-guard/): Security toolkit of input/output scanners for LLM interactions. Repository is archived (read-only). (Open source, Protect AI)
- [NeMo Guardrails](https://agentsecurityindex.com/tools/nemo-guardrails/): Open-source toolkit for adding programmable guardrails (topic, safety, jailbreak rails) to LLM apps. (Open source, NVIDIA)
- [NeuralTrust](https://agentsecurityindex.com/tools/neuraltrust/): AI gateway and guardrails that protect agents and apps from attacks, hallucinations and data leaks. (Commercial, NeuralTrust)
- [Pangea AI Guard](https://agentsecurityindex.com/tools/pangea-ai-guard/): API services for securing AI apps, including prompt-layer guardrails and data redaction. (Commercial, Pangea)
- [Prompt Security](https://agentsecurityindex.com/tools/prompt-security/): Security platform for employee and homegrown GenAI use: prompt protection, data leak prevention and AI usage visibility. (Commercial, Prompt Security)
- [PurpleLlama (Llama Guard, Prompt Guard, LlamaFirewall)](https://agentsecurityindex.com/tools/purplellama/): Meta's set of open tools to assess and improve LLM security, including Llama Guard classifiers and LlamaFirewall. (Open source, Meta)
- [Vigil](https://agentsecurityindex.com/tools/vigil/): Python library and API that scans LLM inputs for prompt injection and jailbreaks. Last code push Jan 2024. (Open source, Independent (deadbits))

## MCP gateways & scanners

- [agentgateway](https://agentsecurityindex.com/tools/agentgateway/): Open-source data plane for LLM, MCP, A2A and HTTP traffic. (Open source, agentgateway project)
- [Cisco MCP Scanner](https://agentsecurityindex.com/tools/cisco-mcp-scanner/): Scans MCP servers for potential threats and security findings. (Open source, Cisco AI Defense)
- [ContextForge MCP Gateway](https://agentsecurityindex.com/tools/contextforge-mcp-gateway/): Gateway, registry and proxy in front of MCP, A2A and REST/gRPC APIs with guardrails and central discovery. (Open source, IBM)
- [Docker MCP Gateway](https://agentsecurityindex.com/tools/docker-mcp-gateway/): Docker CLI plugin and gateway for running and managing containerized MCP servers. (Open source, Docker)
- [Kong AI Gateway](https://agentsecurityindex.com/tools/kong-ai-gateway/): AI gateway adding security, routing, observability and cost control for LLM and MCP traffic. (Open source + commercial, Kong)
- [Lasso MCP Gateway](https://agentsecurityindex.com/tools/lasso-mcp-gateway/): Plugin-based gateway that orchestrates other MCP servers with security plugins. (Open source, Lasso Security)
- [Lunar.dev MCPX](https://agentsecurityindex.com/tools/lunar-dev-mcpx/): Enterprise AI control plane for MCP, LLM and API governance with observability. (Commercial, Lunar.dev)
- [MCP-Shield](https://agentsecurityindex.com/tools/mcp-shield/): Security scanner for MCP servers. Last code push Apr 2025. (Open source, Independent)
- [Microsoft MCP Gateway](https://agentsecurityindex.com/tools/microsoft-mcp-gateway/): Reverse proxy and management layer for MCP servers with session-aware routing on Kubernetes. (Open source, Microsoft)
- [MintMCP](https://agentsecurityindex.com/tools/mintmcp/): Managed MCP gateway that gives teams governed access to models, MCP tools and context. (Commercial, MintMCP)
- [Obot MCP Gateway](https://agentsecurityindex.com/tools/obot-mcp-gateway/): Enterprise AI control plane with an MCP gateway for deploying and governing agents. (Open source + commercial, Obot)
- [Runlayer](https://agentsecurityindex.com/tools/runlayer/): Platform for AI enablement, security and control over agents and MCP. (Commercial, Runlayer)
- [Snyk Agent Scan (formerly mcp-scan)](https://agentsecurityindex.com/tools/snyk-agent-scan/): Scanner for AI agents, MCP servers and agent skills; originated as Invariant Labs' mcp-scan. (Open source, Snyk)

## Agent runtime security & posture

- [Cisco AI Defense](https://agentsecurityindex.com/tools/cisco-ai-defense/): Discovers AI assets, red teams models and enforces runtime guardrails from the network layer. (Commercial, Cisco)
- [HiddenLayer](https://agentsecurityindex.com/tools/hiddenlayer/): End-to-end AI security platform that detects threats against models and agents at runtime. (Commercial, HiddenLayer)
- [Knostic](https://agentsecurityindex.com/tools/knostic/): Secures AI coding agents (Claude Code, MCP servers, skills, IDE extensions) used inside a company. (Commercial, Knostic)
- [Noma Security](https://agentsecurityindex.com/tools/noma-security/): Agent security and governance platform for enterprise security teams. (Commercial, Noma Security)
- [Operant AI](https://agentsecurityindex.com/tools/operant-ai/): Real-time discovery, detection and defense for AI apps, agents and MCP. (Commercial, Operant AI)
- [Pillar Security](https://agentsecurityindex.com/tools/pillar-security/): Platform to discover, govern and secure every AI agent and workflow across an organization. (Commercial, Pillar Security)
- [Prisma AIRS](https://agentsecurityindex.com/tools/prisma-airs/): Palo Alto Networks' AI security platform to discover, assess and protect AI apps, models, agents and data. (Commercial, Palo Alto Networks)
- [Straiker](https://agentsecurityindex.com/tools/straiker/): Detects prompt injection, tool misuse and runtime attacks across agentic AI apps. (Commercial, Straiker)
- [WitnessAI](https://agentsecurityindex.com/tools/witnessai/): AI security and governance with network visibility, intent-based controls and runtime protection. (Commercial, WitnessAI)
- [Wiz AI-SPM](https://agentsecurityindex.com/tools/wiz-ai-spm/): AI security posture management: visibility and risk reduction across models, training data and AI services in the cloud. (Commercial, Wiz)
- [Zenity](https://agentsecurityindex.com/tools/zenity/): Evaluates every AI agent action in real time to control what agents can and cannot do. (Commercial, Zenity)

## AI red teaming & evals

- [Adversa AI](https://agentsecurityindex.com/tools/adversa-ai/): AI red teaming and protection, including monitoring of what coding agents do. (Commercial, Adversa AI)
- [AgentDojo](https://agentsecurityindex.com/tools/agentdojo/): Research benchmark environment for evaluating prompt-injection attacks and defenses on LLM agents. (Open source, ETH Zurich SPY Lab)
- [DeepTeam](https://agentsecurityindex.com/tools/deepteam/): Open-source framework to red team LLMs and AI agents. (Open source, Confident AI)
- [Enkrypt AI](https://agentsecurityindex.com/tools/enkrypt-ai/): Red teaming, guardrails and monitoring platform for AI security and safety. (Commercial, Enkrypt AI)
- [FuzzyAI](https://agentsecurityindex.com/tools/fuzzyai/): Automated LLM fuzzing tool to find jailbreaks in LLM APIs. (Open source, CyberArk)
- [garak](https://agentsecurityindex.com/tools/garak/): LLM vulnerability scanner that probes models for jailbreaks, injection, leakage and more. (Open source, NVIDIA)
- [Giskard](https://agentsecurityindex.com/tools/giskard/): Continuous red teaming for AI agents, plus an open-source testing library. (Open source + commercial, Giskard)
- [Gray Swan](https://agentsecurityindex.com/tools/gray-swan/): Security tools and red-teaming for LLMs from researchers behind major AI vulnerability work. (Commercial, Gray Swan)
- [Inspect](https://agentsecurityindex.com/tools/inspect/): Open-source framework for LLM and agent evaluations from the UK AI Security Institute. (Open source, UK AI Security Institute)
- [Mindgard](https://agentsecurityindex.com/tools/mindgard/): Discovers, assesses and red teams AI models, agents and applications. (Commercial, Mindgard)
- [Promptfoo](https://agentsecurityindex.com/tools/promptfoo/): Open-source CLI and platform for evals, red teaming and vulnerability scanning of LLM apps. (Open source + commercial, Promptfoo)
- [promptmap](https://agentsecurityindex.com/tools/promptmap/): Prompt-injection scanner for custom LLM applications. (Open source, Independent (utkusen))
- [PyRIT](https://agentsecurityindex.com/tools/pyrit/): Microsoft's Python Risk Identification Tool for red teaming generative AI systems. (Open source, Microsoft)
- [SPLX](https://agentsecurityindex.com/tools/splx/): AI security testing and red-teaming platform for AI assistants and agents. (Commercial, SPLX)

## Agent identity & authorization

- [Aembit](https://agentsecurityindex.com/tools/aembit/): Runtime IAM for AI agents and workloads that keeps downstream credentials out of agents. (Commercial, Aembit)
- [Arcade.dev](https://agentsecurityindex.com/tools/arcade-dev/): Actions runtime that handles auth and governs every tool call an enterprise agent makes. (Commercial, Arcade.dev)
- [Astrix Security](https://agentsecurityindex.com/tools/astrix-security/): Visibility and governance for AI agent and non-human identities. Site says standalone sales of new licenses have ended. (Commercial, Astrix Security)
- [Auth0 for AI Agents](https://agentsecurityindex.com/tools/auth0-for-ai-agents/): Auth0 identity features for shipping AI agents securely, with user login and delegated access to APIs. (Commercial, Okta (Auth0))
- [Cerbos](https://agentsecurityindex.com/tools/cerbos/): Fine-grained, contextual authorization for apps and AI systems; open-source policy engine. (Open source + commercial, Cerbos)
- [Clutch Security](https://agentsecurityindex.com/tools/clutch-security/): Visibility, governance and zero trust for non-human identities, AI agents and secrets. (Commercial, Clutch Security)
- [Descope](https://agentsecurityindex.com/tools/descope/): Identity platform covering users, partners, AI agents and MCP servers. (Commercial, Descope)
- [Keycard](https://agentsecurityindex.com/tools/keycard/): Runtime authorization that gives each agent its own identity and scoped access to resources. (Commercial, Keycard)
- [Oasis Security](https://agentsecurityindex.com/tools/oasis-security/): Lifecycle management and security for non-human identities and AI agents. (Commercial, Oasis Security)
- [Oso](https://agentsecurityindex.com/tools/oso/): Monitoring, access controls and automated guardrails for running agents in production. (Commercial, Oso)
- [Permit.io](https://agentsecurityindex.com/tools/permit-io/): Fine-grained authorization as a service (RBAC, ABAC, ReBAC) for apps, APIs and AI agents. (Commercial, Permit.io)
- [Stytch](https://agentsecurityindex.com/tools/stytch/): Authentication and security APIs and SDKs for apps. (Commercial, Stytch)
- [Teleport](https://agentsecurityindex.com/tools/teleport/): Unified identity and access for humans, machines and AI agents across infrastructure. (Open source + commercial, Teleport)

## Agent audit & observability

- [AgentOps](https://agentsecurityindex.com/tools/agentops/): Tracing, debugging and replay for AI agents. (Open source + commercial, AgentOps)
- [Arize Phoenix](https://agentsecurityindex.com/tools/arize-phoenix/): Self-hostable tracing, evaluation and debugging for agents and LLM apps. (Open source + commercial, Arize AI)
- [Datadog LLM Observability](https://agentsecurityindex.com/tools/datadog-llm-observability/): Tracing, evaluation and experimentation for AI agents inside Datadog. (Commercial, Datadog)
- [Galileo](https://agentsecurityindex.com/tools/galileo/): AI evaluation, observability and real-time guardrails for agents; now part of Splunk Observability. (Commercial, Galileo)
- [Helicone](https://agentsecurityindex.com/tools/helicone/): Open-source LLM gateway with routing, logging and monitoring. (Open source + commercial, Helicone)
- [Laminar](https://agentsecurityindex.com/tools/laminar/): Open-source platform to trace, evaluate and debug AI agents. (Open source + commercial, Laminar)
- [Langfuse](https://agentsecurityindex.com/tools/langfuse/): Open platform to trace, evaluate and improve AI agents using production data. (Open source + commercial, Langfuse)
- [LangSmith](https://agentsecurityindex.com/tools/langsmith/): Agent and LLM observability with tracing and real-time monitoring. (Commercial, LangChain)
- [OpenLLMetry](https://agentsecurityindex.com/tools/openllmetry/): OpenTelemetry-based open-source instrumentation for GenAI and LLM apps. (Open source, Traceloop)
- [Opik](https://agentsecurityindex.com/tools/opik/): Open-source AI observability to test, trace and monitor LLM apps and agents. (Open source + commercial, Comet)
- [W&B Weave](https://agentsecurityindex.com/tools/w-b-weave/): Tracing and evaluation toolkit for LLM and agent applications. (Open source + commercial, Weights & Biases)

## About

- [Methodology](https://agentsecurityindex.com/methodology/): how tools are listed, verified and lab-tested
- [About](https://agentsecurityindex.com/about/): who runs Agent Security Index, corrections policy
- [Disclosure](https://agentsecurityindex.com/disclosure/): how the site makes money and stays independent
- [Changelog](https://agentsecurityindex.com/changelog/): log of verification passes
- [For vendors](https://agentsecurityindex.com/get-featured/): featured listings, newsletter sponsorship, independent lab tests
- Contact: hello@agentsecurityindex.com