AI agent security and governance for the enterprise
Zenity is an AI agent security and governance platform for enterprises operating across SaaS, cloud, and endpoint environments.
AI Panel Score
6 AI reviews
Reviewed
AI Editor ApprovedApproved and published by our AI Editor-in-Chief after full panel analysis.In practice, Zenity connects to an organization's SaaS platforms, cloud environments, and endpoints to build an inventory of AI agents, copilots, and automations already in use, including ones created outside formal IT channels. Security and risk teams use this inventory to assess posture, identify misconfigurations or excessive permissions, and monitor agent behavior for signs of misuse or compromise. When threats are detected, the platform supports inline prevention to block or contain risky agent actions in real time rather than only alerting after the fact.
The platform is organized around named use cases tied to specific systems where agentic AI is deployed, including Microsoft Copilot Studio, Microsoft Foundry, Salesforce and Salesforce Agentforce, ServiceNow, and ChatGPT Enterprise, as well as broader risk categories such as MCP (Model Context Protocol) implementations. Zenity also maintains a research arm, Zenity Labs, which has published findings on agent vulnerabilities (such as the AgentFlayer research) and collaborates with frameworks like MITRE ATLAS on classifying AI agent attack techniques. The company describes a concept of "guardian agents" as part of its approach to agentic AI governance.
Establishes a framework of guardian AI agents designed to monitor and protect other AI agents within an organization.
Discovers AI agents, copilots, and automations across an organization's SaaS, cloud, and endpoint environments.
Secures AI agents across their entire lifecycle, from development/buildtime through live runtime operation.
Delivers security oversight and governance for organizations using ChatGPT Enterprise.
Provides dedicated security and governance coverage for AI agents built and deployed in Microsoft Copilot Studio.
Secures AI agents deployed within Salesforce Agentforce environments.
Integrates with ServiceNow to secure AI and automation workflows running on the platform, including SecOps use cases.
Provides AI-driven incident detection and response capabilities for identifying and reacting to threats against AI agents.
Applies inline controls to block risky or malicious AI agent actions before they execute.
Addresses security risks associated with Managed Cloud Platforms (MCP) used in agentic AI deployments.
Assesses and manages the security posture of discovered AI agents and automations across the enterprise.
Monitors AI agent activity to detect real-time threats and malicious behavior.
Zenity is an enterprise AI agent security and governance platform sold via a sales-led model; no public list pricing is available and organizations must contact Zenity directly for a custom quote based on scale and needs.
Real problem, real research, a $125M round behind it — and pricing still hidden in a sales call.
“Zenity covers a gap most CISOs don't know they have yet, and a $125M Series C says the market agrees. Sales-led pricing with no public number is still the harder half of the board conversation.”
Copilot Studio, Salesforce Agentforce, ServiceNow, ChatGPT Enterprise. That's the right list of integrations today. Every enterprise has shadow agents already; the question is whether anyone's watching them.
Zenity Labs publishing AgentFlayer research and working with MITRE ATLAS is the kind of signal that matters more than a feature list. It says the team understands attacks, not just dashboards. Compare that to CrowdStrike or Microsoft's own Defender bolting on agent visibility as a checkbox — Zenity's built for this specific problem.
Two things worry me. No public pricing means every deal is a negotiation, and three published case studies — Varonis and Telit Cinterion among them — is a short list to hand a board asking who else runs this at scale. Pilot it against one high-risk surface — Copilot Studio or ChatGPT Enterprise — before any enterprise-wide commitment.
Few dedicated players cover Copilot Studio and Agentforce natively; adopting early is a differentiator versus peers still relying on generic DLP tools.
Being early on agent security looks smart if it works, sketchy if the vendor can't produce reference customers on request.
Discovery of shadow agents (per the platform's core capability) can show value in weeks, but inline prevention rollout across ServiceNow, Salesforce, and Microsoft stacks is a longer integration lift.
Full buildtime-to-runtime lifecycle coverage addresses a genuinely new risk category, not just cheaper monitoring of known threats.
A $125M Series C led by Norwest in August 2026 and 230-plus staff answer the three-year question; the open risk is the category, not the balance sheet.
Enterprises already running Copilot Studio, Agentforce, or ChatGPT Enterprise at scale and worried about shadow agents.
Skip if your agent footprint is still small or experimental and a generic CASB covers your current risk.
Shadow AI agents are my new shadow IT problem, and Zenity is built for exactly that gap.
“Zenity attacks the unmanaged-agent inventory problem that no CASB or CSPM tool was designed to catch. It's an emerging-category bet, not a mature one, but the lifecycle coverage and named integrations are the right architecture.”
Unmanaged AI agents are today's version of shadow SaaS, except these things can take autonomous action with production credentials. Zenity's discovery-first model, mapping agents across Microsoft Copilot Studio, Salesforce Agentforce, ServiceNow, and ChatGPT Enterprise, is the correct starting posture. I can't govern what I can't inventory, and agent sprawl outside formal IT channels is precisely where my next incident originates.
Intent-based detection examining tool calls, memory access, and control flow, rather than prompt-string filtering, tells me their research arm (Zenity Labs, AgentFlayer) is doing real adversarial work, not marketing. MITRE ATLAS and OWASP mapping gives my audit team a defensible framework.
The tradeoff: no public pricing, sales-led motion, and zero free trial means a long procurement cycle before I can validate inline prevention against my actual agent fleet. Competing against CrowdStrike and Microsoft's native Purview extensions for budget won't be trivial, but the buildtime-to-runtime scope is more complete than either.
Sits ahead of generic CASB/CSPM tools by treating agents as a first-class asset class, competing directly with CrowdStrike's emerging AI-SPM push.
Discovery-plus-posture-plus-runtime lifecycle mirrors how CISOs actually triage unmanaged shadow-IT-style risk.
SaaS, cloud and endpoint coverage claimed, and docs.zenity.io publishes the per-integration setup — Azure, Salesforce, ServiceNow, Bedrock — openly enough to validate before a purchase.
Named integrations (Copilot Studio, Agentforce, ServiceNow) create dependency on Zenity tracking each vendor's agent framework changes.
Intent-based detection across tool calls, memory, and control flow goes beyond prompt-firewall approaches most competitors ship.
Enterprises with active Microsoft Copilot Studio, Salesforce Agentforce, or ServiceNow agent deployments needing lifecycle-wide governance now.
Avoid if your organization has minimal agentic AI footprint or needs self-serve pricing and a fast proof-of-concept without a sales cycle.
Zero dollars on the page. Zero tiers. Zero comparison math possible.
“No published pricing means no TCO model, just a sales call. Enterprise security budgets deserve better than 'contact us.'”
No pricing page. No tiers. No starting price. Just 'contact sales.' That's category norm for enterprise security — Wiz and CrowdStrike play the same game — but it doesn't help finance.
Can't build a 3-year TCO without a number to start from. Guardian Agents, AIDR, MCP risk coverage — real features, zero list price attached. Budget owners get quoted per-seat or per-agent, likely scaled to org size. No way to model seat creep or renewal risk sight unseen.
No free trial, no free plan. Procurement has to run a full sales cycle before seeing a contract. That's fine for a $500K ARR enterprise deal. It's friction for anyone hoping to benchmark against Microsoft Purview or CrowdStrike's agent security add-ons before committing budget.
Sales-led onboarding only, no self-serve path, no invoicing detail available.
No public terms on renewal, term length, or exit — standard enterprise opacity, category norm.
No pricing page, no tiers, no starting price — fully sales-gated.
Intent-based detection and inline prevention offer measurable stops (blocked actions), but no benchmark metrics published.
Feature set (buildtime-to-runtime, four named integrations) suggests real deployment cost but no number to model against.
Enterprises with existing security budget and a procurement team ready for a sales-led evaluation.
You need a number today to compare against CrowdStrike or Microsoft before your board meeting.
Shadow-agent discovery is real, but you're triaging alerts sight-unseen until sales calls you back
“Zenity's pitch — inventory every Copilot Studio flow and Agentforce bot before someone else finds it first — hits a real gap. Docs are open, but no pricing page and no free trial means you're evaluating this on paper until procurement is already in motion.”
Intent-based detection tracking tool calls, memory access, and control flow instead of just prompt content is the right instinct — prompt firewalls alone miss agent logic abuse, and Zenity's own materials call that out directly against a real failure mode. Guardian agents watching other agents is a defensible model for runtime containment, closer to how CrowdStrike frames EDR than how CASB tools frame SaaS visibility.
Day-3 reality is better than the marketing site suggests. docs.zenity.io is open — per-integration setup for Azure, Salesforce, ServiceNow and Bedrock, a Boundaries CLI, release notes. What I couldn't find is a full API reference, and policy-as-code examples would carry more weight than connector checklists.
No free trial, no public pricing, sales-led only. Fine for a security budget line item, bad for a hands-on eval where you'd normally spin up a sandbox and poke at the MCP risk coverage yourself before committing headcount to onboarding it.
Strong detection concept, and docs.zenity.io publishes per-platform setup plus release notes, though real usage patterns still need a live tenant.
docs.zenity.io is open and engineer-shaped — SSO setup, service-principal and connected-app guides, a Boundaries CLI, release notes.
No self-serve trial or pricing page adds procurement friction before any technical friction is even reachable.
MITRE ATLAS mapping and Zenity Labs research (AgentFlayer) signal depth beyond checkbox compliance.
Named integrations for Copilot Studio, Agentforce, ServiceNow, ChatGPT Enterprise map directly onto where agents actually live.
Enterprise security teams already running Copilot Studio or Agentforce at scale who need shadow-agent discovery now.
Avoid if you need a self-serve trial or public docs before looping in procurement.
Serious security tool for a problem most companies don't know they have yet.
“Zenity watches the AI agents your employees spun up without telling IT. Whether that's a daily win or a quarterly report depends entirely on who's using it.”
No pricing page, no free trial, straight to 'contact sales.' That's normal for enterprise security, but it means I can't tell you what day one feels like — nobody outside a sales call has. What I can tell from the site is the feature list is real, not vaporware: Copilot Studio, Salesforce Agentforce, ServiceNow, ChatGPT Enterprise, all named integrations, plus intent-based detection that traces tool calls and memory access instead of just scanning prompts like a firewall would.
This isn't a tool you open at 9am. It's a tool your SOC team lives inside of, or ignores until an incident happens. That's the whole category though — compare it to how Wiz approaches cloud posture, alert fatigue is the real enemy, not the UI.
The guardian agents concept and Zenity Labs research (AgentFlayer) suggest real technical depth. Whether day-to-day dashboards feel like homework, nobody outside a pilot knows yet.
Release notes are published, but I couldn't find product screenshots, so daily UI quality stays unverifiable either way.
Named use cases per platform (Copilot Studio, Agentforce, ServiceNow) help focus onboarding but the surface area is genuinely large.
It's a web console as far as I can tell; this is a SOC category so mobile isn't really the point, but there's nothing for a phone.
Sales-led with no free trial means onboarding is a negotiated deployment, not a self-serve first hour.
Inline prevention and runtime detection mapped to OWASP/MITRE ATLAS suggest a mature detection pipeline, not just alerts.
Enterprise security teams already juggling Copilot Studio, Agentforce, or ServiceNow agents built outside IT's knowledge.
You want to try before you buy or you're a smaller org without a dedicated SOC team.
No public pricing, sales-led, and 'Guardian Agents' branding on a category still forming.
“Zenity's coverage list reads well — Copilot Studio, Agentforce, ServiceNow, ChatGPT Enterprise. But pricing isn't public anywhere, and 'Guardian Agents' is the kind of framing that needs a lot more proof than a landing page gives.”
'Guardian agents' as a category name. I've heard variations of this from CASB vendors circa 2015 and CSPM vendors circa 2019 — 'we'll watch the thing watching your stuff' is a durable pitch, not always a durable company. Netskope and Wiz both survived similar framing. Plenty of others didn't.
What's real here: named integrations (Copilot Studio, Agentforce, ServiceNow, ChatGPT Enterprise), a research arm publishing actual findings (AgentFlayer), and MITRE ATLAS alignment. That's more grounding than most AI-security decks show. Docs are open at docs.zenity.io, release notes and all, so shipping cadence is checkable from outside.
Contact-sales-only pricing at this stage is category norm, not a red flag by itself. Exit portability is the real question: agent governance data and policies built up over 18 months don't travel light if you switch vendors. Competitive field is thin but filling fast — CrowdStrike and Palo Alto both circling this space.
Multi-platform breadth (Copilot Studio, Agentforce, ServiceNow, ChatGPT Enterprise) is wider than most single-platform AI security tools today.
Docs are open at docs.zenity.io, but I couldn't find an export path; posture and policy data built up over time likely doesn't travel.
A $125M Series C in August 2026, 230-plus staff, and active Zenity Labs research answer the runway question; the category question stays open.
Claims are specific (named platforms, MITRE ATLAS mapping) rather than pure superlative, but 'guardian agents' framing is aspirational.
Fits the CASB/CSPM discovery-then-control playbook that has both winners and graveyard entries.
Enterprises already running Copilot Studio, Agentforce, or ChatGPT Enterprise at scale who need agent inventory now.
You need transparent self-serve pricing or want to evaluate without a sales cycle.
Common questions answered by our AI research team
Zenity offers dedicated integrations for Microsoft Copilot Studio, Salesforce Agentforce, ServiceNow, and ChatGPT Enterprise, covering AI agents and copilots across SaaS, cloud, and endpoint environments.
Yes, Zenity has a dedicated integration for Microsoft Copilot Studio as part of its coverage across leading agent-building platforms.
Zenity uses intent-based detection, examining the full execution path including tool calls, memory access, data usage, and control flow to identify malicious or unintended outcomes even when inputs look harmless. This approach exposes attacks that prompt-based firewalls miss and correlates configuration, permissions, and runtime behavior into actionable threat signals.
Zenity covers the full agent lifecycle from buildtime to runtime, combining preemptive risk prevention during configuration with runtime threat detection, investigation, and response mapped to OWASP and MITRE ATLAS.
Yes, Zenity provides unified visibility, policy control, and threat prevention across SaaS, home-grown agentic platforms (Cloud), and end-user devices (Endpoint).
Who Zenity competes with, and how each one differs
The closest pure-play AI security peer: both discover AI assets and add runtime protection. HiddenLayer centres on the model layer — model scanning, AI supply-chain security and adversarial-ML attack simulation — while Zenity works at the agent and copilot application layer inside Microsoft, Salesforce, OpenAI and AWS platforms, governing what deployed agents are allowed to do.
The most direct head-to-head rival: another Israeli-founded AI agent security platform selling AI security posture management plus runtime guardrails to enterprise CISOs, and it raised $100M in July 2025 to chase the same buyers. Noma leans more toward the AI/ML development pipeline and self-built agents; Zenity leans toward agents built on packaged SaaS platforms such as Copilot Studio and Agentforce.
Attacks the same problem as a platform incumbent through Prisma AIRS, which added agentic AI protection in 3.0, backed by the Protect AI and Portkey acquisitions. It bundles AI security into an enterprise-wide network and cloud security contract, whereas Zenity sells as a standalone AI-agent-specific product that can be bought without displacing an existing security stack.
Entered enterprise AI security by acquiring Pangea for $260M in September 2025, extending Falcon into AI detection and response for employee and agent use of GenAI. CrowdStrike leads with endpoint and identity telemetry and an existing agent on the box; Zenity is agentless and derives its signal from the SaaS and low-code platforms where business agents are actually built.
Bought Prompt Security in August 2025 for roughly $250M to add GenAI and copilot security to its XDR platform, overlapping Zenity on prompt-injection defence and shadow-AI discovery. SentinelOne's route to market is the existing endpoint and SIEM footprint; Zenity competes on depth of coverage for agent build-time posture and per-agent policy rather than breadth of platform.
Zenity provides a security and governance platform for AI agents, helping enterprises manage risk from buildtime to runtime. It is based in Boston, MA.
