AI Agent Security: 11 Tools compared

Harden's Agentic Integrity Foundation is a guardrail for AI coding agents that runs on the developer's own machine. It checks each supported tool call an agent is about to make before it executes, and then allows it, redacts it, asks, logs it, or blocks it and suggests a safer retry. Harden says the checks run against its own post-trained model using the request and session context, and that repository content and tool output stay on the machine. It attaches to agents through native hooks where those exist and through an MCP proxy for the rest, and the vendor lists Claude Code, Codex, Cursor, Antigravity CLI, Kiro, Hermes and OpenClaw as supported, with a version matrix it rechecks on each release. Decisions are kept in a local history. The full local model needs macOS on Apple Silicon, the command line tool and daemon run on macOS or Linux x86_64, at least 16 GB of memory is required and Windows is not supported. Harden publishes its own results for four agent-security benchmarks against a GPT monitor baseline. The company was founded by Pushpak Pujari and Pushpendre Rastogi.

HOL Guard is an open-source, local-first runtime security layer that sits between an AI coding agent and the tools it wants to run. It evaluates shell commands, file access, package installs and MCP tool calls before execution and allows, blocks or routes them for approval according to policy, with the vendor stating that checks take under 50 ms and that files never leave the machine. It supports agents including Claude Code, Codex, Cursor, Gemini CLI, OpenCode and GitHub Copilot CLI, installs via pipx on Python 3.10 or later, and offers desktop installers for Apple silicon macOS 13 and later, Linux x86_64 and Windows x64, with Windows marked experimental. The core is Apache-2.0 and needs no account, while an optional Guard Cloud adds shared history, team policy and dashboards at USD 4.99 (Solo), USD 15 (Pro) and USD 30 per seat (Team) per month. It is made by HOL, operated by Hashgraph Online DAO LLC, a consortium established in November 2024 whose founding members include HashPack and Hgraph. The GitHub repository shows 570 stars and 3,705 commits, and PyPI lists version 3.0.131 released on 8 September 2026.

Lasso Security is an AI security platform for organisations that build or deploy AI agents and LLM applications. Its modules cover discovery and an AI bill of materials, AI security posture management, automated AI red teaming with a library of more than 3,000 attacks, runtime enforcement and AI detection and response, and its agent policies define what tools agents can use and which resources they are permitted to access. It deploys via gateway, API or SDK, with a browser extension for employee AI usage, and the company publishes an open-source MCP Gateway under the MIT licence. The company was founded in 2023, is headquartered in Tel Aviv with an office in New York, and announced a $30 million round led by ClearSky in September 2026, taking total funding past $37 million. It is SOC 2 Type II certified, and its AWS Marketplace SaaS listing for the Secured Gateway for LLMs shows $50,000 for a 12-month contract.

Pillar Security is a platform to discover, test, secure and govern AI agents across an organisation. Its four modules are AI Discovery and Posture (cataloguing agents, models, prompts, frameworks, tools, MCP servers and skills), Agentic Red Teaming (multi-turn adversarial attacks against complete AI systems), Runtime Guardrails (malicious intent detection, agent behaviour analysis and masking of PII, PHI and credentials) and Governance and Compliance. It integrates with code and CI/CD, AI gateways and orchestration, MCP servers and skills marketplaces, and models and AI platforms, and can be deployed on the customer's own VPC with their own data and models. The company was founded in late 2023 by Dor Sarig (CEO) and Ziv Karliner (CTO), is based in Tel Aviv and Miami, and announced a $9 million seed round led by Shield Capital in April 2025; Refresh Miami reported around 20 employees in January 2026. The site states a SOC 2 Type II audit, and no pricing is published.

ranwhat is an open-source command-line tool that reviews what AI coding agents have already done on a developer's machine. It reads the history that Claude Code, Codex, Gemini CLI, GitHub Copilot CLI and eight other agents write to disk, and picks out the irreversible actions worth knowing about, such as reading private keys, recursive deletions and publishing packages. A clean command finds credentials left in plaintext in agent transcripts, says which to rotate and masks them on request, and a scan command scores the authority an agent's credentials carry, including permissions that were granted but never used. It works after the fact rather than blocking actions as the agent runs, and the site points to the agents' own permission rules, hooks and sandboxes for that. The tool is written in Python, MIT-licensed and published on PyPI, and the site states that it sends no telemetry and needs no account. A paid Plus plan adds a capability catalogue feed now, with shared reporting, retention and alerts listed as still being built. It was first released in September 2026 by Matija Mikulinec.

Zenity is an AI agent security and governance platform organised into three layers: Surface, which builds a live inventory of agents across SaaS, custom and endpoint deployments and validates exploitable attack paths; Enforce, which evaluates and blocks unsafe agent actions before they execute; and Protect, which detects runtime threats and correlates them against OWASP and MITRE ATLAS. Its documentation lists integrations for Microsoft Copilot Studio, Power Platform, Azure AI Foundry, ChatGPT Enterprise, Salesforce Agentforce, ServiceNow, Amazon Bedrock, Google Vertex AI and Gemini Enterprise, Anthropic, LiteLLM, GitHub Copilot and Cursor, and the company describes native hooks and a Zenity MCP gateway for coding agents such as Claude Code, Codex, GitHub Copilot and Cursor. The platform is SaaS; on AWS Marketplace it is priced per resource and per million tokens with a 30-day free trial, and it is also listed on the Microsoft Azure Marketplace. Zenity was founded in 2021 by Ben Kliger and Michael Bargury, runs R&D from Tel Aviv and go-to-market from New York, and announced a USD 125 million Series C led by Norwest on 3 August 2026, bringing total funding to about USD 185 million according to Calcalist. It remains independent. The trust centre lists SOC 2 Type II, ISO 27001 and ISO 27701.