DeepKeep
Verified Oct 16, 2026 ActiveRuntime firewall, red teaming and model scanning for LLM, vision and agent systems in production
Listed on EffectHub ·
Pricing
- Pricing model
- Contact sales
No price on the site; every platform call to action leads to a demo booking, which points to enterprise quoting. A separate free AI Agent Scanner offers a one-off agentic attack-surface assessment.
Use tool ↗About DeepKeep
DeepKeep secures the AI layer rather than the infrastructure beneath it. The platform pairs a runtime firewall — more than 60 guardrails whose action can be set per policy to block, redact, replace or alert — with adversarial red teaming, static and dynamic model scanning for supply-chain risk, an agent scanner that maps an agent's tools and attack surface, and a visibility layer for governing what employees feed into AI tools. It covers text, image and video models and deploys as SaaS, on-premises, in a private VPC or fully air-gapped. The vendor is DeepKeep Ltd., based in Tel Aviv and founded by Rony Ohayon; the site lists a granted US patent plus ISO 27001, ISO 9001, SOC 2 and GDPR compliance.
- Platforms
- Web · API
- API
- Yes
- Last verified
- Oct 16, 2026
Verified sources
- Home page lists AI Firewall, AI Red Teaming (new Vibe AI Red Teaming), AI Lens, AI Agent Scanner and Model Scanning for LLM and vision; all calls to action are Book a Demo, no pricing page (/pricing returns 404); footer © 2026 DeepKeep (15 October 2026)deepkeep.ai ↗
- About page lists the leadership team and DeepKeep patent US-12450363-B2 (15 October 2026)deepkeep.ai ↗
- Free AI Agent Scanner offers an agentic attack-surface assessment (15 October 2026)agentscanner.deepkeep.ai ↗
Is DeepKeep legit?
We don’t label tools safe or unsafe. Below are the records we checked ourselves, each with the date we read it.
- Domain registered
- Mar 23, 2021 (5 years ago)Registry record (RDAP), registrar GoDaddy.com, LLC, looked up Sep 18, 2026
- TLS certificate
- Issued by Google Trust Services, valid until Dec 31, 2026Seen in our own connection to the site on Oct 16, 2026
- Last checked
- Oct 16, 2026The site answered
- Operated by
- DeepKeep Ltd., IsraelOur research, card verified Oct 15, 2026
Pros & cons of DeepKeep
Broader than the usual prompt-injection filter: alongside runtime guardrails it does model supply-chain scanning, adversarial testing and an agent scanner that maps what tools an agent can actually reach. Vision and video coverage is a real differentiator against text-only competitors, and air-gapped deployment opens regulated buyers. The cost is total pricing opacity and a surface area wide enough that depth will vary by module.
Pros
- Firewall actions are policy-configurable — block, redact, replace or alert — rather than a single verdict
- Covers vision and video models alongside text LLMs
- Agent scanner maps an agent's tools and attack surface, which most guardrail vendors skip
- Deploys air-gapped or in a private VPC for regulated environments
- Documented API and integration guides on a separate docs site
Cons
- No pricing anywhere; every path ends at a demo booking
- Five distinct capabilities marketed as one platform, which usually means uneven depth
- Enterprise-only motion with no self-serve trial
- Public evidence of scale is limited to logo walls and analyst mentions
Best for
- Enterprises putting agents or LLM apps into production under audit
- Regulated buyers who need on-premises or air-gapped deployment
- Teams securing vision and video models, not only text
Not for
- Small teams wanting a self-serve guardrail library
- Buyers who need a published price to start a procurement
Watch out: Firewall, red teaming, model scanning, agent scanner and usage control are separate products under one platform name — check which are actually in a quote.