DeepKeep
by DeepKeep Ltd.
AI security platform covering red teaming, an AI firewall, agent attack-surface scanning and runtime controls for coding agents
DeepKeep is an AI security and trust platform that protects LLM, computer-vision and agentic applications across their lifecycle with red teaming, an AI firewall, model scanning and usage controls. It is aimed at CISOs and AI platform teams who need to secure both the AI their company builds and the AI tools employees and developers use.
DeepKeep is a Tel Aviv company founded in 2021 by CEO Rony Ohayon, with Yossi Altevet as co-founder and CTO. It came out of stealth in May 2024 with a $10 million seed round led by Awz Ventures and positions itself as an AI-native trust, risk and security management (TRiSM) platform. The product is a set of modules on one policy engine. AI Red Teaming tests foundation models and applications for prompt injection, jailbreaks and data leakage, including a human-steered 'Vibe AI Red Teaming' mode driven by an agent called Reddy. The AI Firewall evaluates prompts and responses at runtime with more than 60 contextual guardrails and is natively multimodal across text, images and video. Model Scanning checks the supply chain with MLBOM and CVE detection. The AI Agent Scanner, launched in March 2026, maps the tools, intents, data sources and vulnerabilities of agentic workflows built on Microsoft frameworks, Salesforce Agentforce, OpenAI Agents, CrewAI, Amazon Bedrock AgentCore, n8n and Make, referencing the OWASP Top 10 for Agentic Applications. AI Lens provides discovery, shadow-AI visibility and role-based usage control; on October 1, 2026 DeepKeep added AI Lens for Developers, a lightweight plug-in that hooks into Cursor and Claude Code to inspect prompts, file reads, shell commands and MCP tool calls, flag leaked credentials and insecure code, route destructive commands for human approval and log every session. It deploys as SaaS, private VPC, on-premises or air-gapped, lists SOC 2, ISO 27001 and GDPR compliance, and names customers including ST Engineering, Macnica and EY. It won The Hacker News 2026 Cybersecurity Stars award for best AI security platform.
A CISO or AI security lead at a regulated enterprise that needs one platform to test its own models and agents, guard them at runtime and control how employees and developers use third-party AI tools.
Red teaming, runtime guardrails, agent attack-surface mapping and coding-agent controls under a single policy engine, deployable even in air-gapped environments.
At a Glance
- Category
- Governance & Security
- Pricing
- Contact for pricing, Freemium
- Target Market
- CISOs, AI Platform Teams, Security Engineers, Enterprise Developers
- Deployment
- Hybrid, Self-hosted
- Founded
- 2021
- Headquarters
- Tel Aviv, Israel
- Team Size
- 51-200
Key Features
- ✓AI Firewall
Evaluates prompts and responses in real time with more than 60 contextual, multilingual guardrails across text, images and video to stop injection and leakage.
- ✓AI Red Teaming
Automated and human-steered adversarial testing of models and applications for prompt injection, jailbreaking and data leakage before they reach production.
- ✓AI Agent Scanner
Maps tools, intents, data sources and vulnerabilities across agent frameworks and produces visual risk maps showing where guardrails should be deployed.
- ✓AI Lens for Developers
A lightweight plug-in that intercepts coding-agent prompts, shell commands, file reads and MCP calls, enforcing allow, block or approval policies.
- ✓Model Scanning
Checks models for supply-chain risk with MLBOM generation and CVE detection, plus pre-deployment assessment of bias, hallucination and leakage.
- ✓Flexible deployment
Runs as SaaS, private VPC, on-premises or air-gapped, so sensitive and classified environments can use the same platform as cloud teams.
Capabilities
Use Cases
- •Governing AI coding agents
Give the security team audit logs and policy control over Cursor and Claude Code, blocking credential leaks and holding destructive commands for approval.
- •Pre-release model red teaming
Test a customer-facing LLM application for jailbreaks and data leakage before launch, then enforce fixes through runtime guardrails.
- •Agent attack-surface mapping
Scan Agentforce, Bedrock AgentCore or n8n workflows to see which tools and data each agent can reach and where exposure is highest.
- •Shadow AI control
Discover which AI tools employees use and apply role-based usage policies so sensitive data does not leave through unsanctioned assistants.
Ideal For
Best For
- ✓Security teams that need visibility and policy control over Cursor and Claude Code on developer machines
- ✓Enterprises red teaming LLM and computer-vision applications before production release
- ✓Organizations mapping the attack surface of agents built on Agentforce, Bedrock AgentCore, CrewAI or n8n
- ✓Defense, government and regulated buyers that require on-premises or air-gapped deployment
Not Ideal For
- ✗Teams standardized on GitHub Copilot, OpenAI Codex or Windsurf that need coding-agent controls today, since support for those agents is still planned
- ✗Buyers who want transparent self-serve pricing, because DeepKeep publishes no price list
- ✗Organizations that prefer a vendor with deep independent practitioner reviews; public user feedback on DeepKeep is still sparse
Deployment
Market Analysis
Pros
- ✓Broad module coverage across the AI lifecycle under one policy engine
- ✓Coding-agent controls arrive as a lightweight plug-in rather than another endpoint agent
- ✓On-premises and air-gapped deployment suits defense and regulated industries
- ✓Third-party recognition including The Hacker News 2026 Cybersecurity Stars award and Gartner and Omdia mentions
Cons
- ✗AI Lens for Developers supports only Cursor and Claude Code at launch; Copilot, Codex and Windsurf are still planned
- ✗Launch coverage gives no data on false-positive rates, performance impact or EDR and SIEM integration
- ✗Only a $10M seed round is publicly disclosed, a small war chest against heavily funded AI security rivals
- ✗No pricing published and little independent practitioner feedback on Hacker News or review sites
Pricing
AI Agent Scanner (free tool)
$0
- ✓Agent attack-surface scan
Enterprise platform
Contact for pricing
- ✓AI Firewall
- ✓AI Red Teaming
- ✓AI Lens and AI Lens for Developers
- ✓Model Scanning
- ✓SaaS, VPC, on-premises or air-gapped
DeepKeep publishes no list pricing; its pricing URL returns 404 and the platform is sold through demos. A free AI Agent Scanner is offered at agentscanner.deepkeep.ai. AI Lens for Developers is included as part of the wider platform rather than sold as a separate SKU, per the launch coverage.
Security & Compliance
Sources
This page was written from 6 sources, 5 on domains other than deepkeep.ai.
- 1.deepkeep.ai — deepkeep.aivendor
- 2.helpnetsecurity.com — deepkeeps ai lens flags coding agent data leaks and routes d
- 3.unite.ai — deepkeep adds runtime guardrails for ai coding agents with a
- 4.siliconangle.com — deepkeep launches ai agent attack surface scanner map enterp
- 5.securityweek.com — deepkeep launches ai native security platform with 10 millio
- 6.awards.thehackernews.com — deepkeep unified ai security
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
Armadin
Autonomous offensive security: AI agent swarms that chain your weak spots into proven attack paths before an attacker does
WitnessAI
Network-layer AI security and governance for employee AI use, models, apps and agents
Ascerta
Enterprise AI management: measure the ROI, cost and adoption of every AI initiative, agent and coding tool
Reco
AI agent security and SaaS security platform that discovers, governs and secures every agent, app and identity