Map / AI Endpoint Security / Harden

Harden

USA · Founded 2025

Quick answer: Harden is an AI security company based in the USA and founded in 2025, working in AI Endpoint Security, Runtime & Guardrails, AI Code Security and 1 more category. It was previously known as Vizops.

AI Endpoint SecurityRuntime & GuardrailsAI Code SecurityObservability & Governance

Capabilities

Agentic Integrity Foundation (AIF): endpoint security for AI coding agents that checks every tool call (shell command, file edit, push, MCP call) at the agent's pre-execution hook, outside the agent harness, against session intent and org policy. Parses the pending command and traces data flow so secrets reaching a network sink are stopped, decodes base64/hex obfuscation, and blocks destructive commands (rm -rf /, force-push to default branch, curl | sh, disk wipes); dynamic code (eval/exec) is routed to a human. Verdicts are allow, block, ask, redact or log-only, judged by post-trained cybersecurity models (~8B) running locally. Separate AI Scanner product does SAST/SCA plus LLM-driven attack-path confirmation and auto-fix PRs for repos, aimed at agent-written code.

Sensors / integration

Local agent on developer macOS/Linux machines (Apple Silicon or x86_64, 16 GB RAM min, no Windows yet) hooking Claude Code, Codex, Cursor IDE/CLI and Antigravity CLI natively, Kiro main agent via hooks with an MCP proxy for subagents, and Hermes/OpenClaw via a bridge process. Free tier installs by curl script with no account; Enterprise deploys via MDM with central policy, SIEM export and air-gapped mode. AI Scanner runs as SaaS or in the customer's AWS/Azure/GCP tenant with CI hooks.

Protections

Blocks, redacts, asks or logs agent actions before execution; fail-closed if the check can't run. Starts in watch-only mode and records every decision.

Differentiator

Sits outside the agent harness with one local policy across all hook-capable coding agents, and positions against both MCP gateways and sandboxes. Pivot: the same founders launched in 2025 as Vizops (vizops.ai), an RL post-training/optimization platform for enterprise AI agents, then moved to AI governance (May 2026) and coding-agent security (AIF, Aug 2026). Founders are from Verkada/AWS (CEO), Google DeepMind/Amazon Alexa (CTO) and WhatsApp/Microsoft (CISO). Pricing: AIF free for individuals, Enterprise custom with a 60-day free trial. The site claims a completed SOC 2 Type II audit, but its trust page still says 'in audit'. No disclosed funding.

Team & funding

Founders
Pushpak Pujari, Pushpendre Rastogi, Anurag Arora
Previously
Vizops

Other AI Endpoint Security companies

Anomity · AstraSync AI · Asymptote Labs · Autonomous · Backslash · Bay Security · Bloom · BlueRock · Bold · Canyon Road · Capsule Security · Certiv · See all →

Last updated · Get the “Listed on” badge