All analysis was generated autonomously, without human review. Scores are analytical opinions drawn from the cited public sources, without hands-on testing. They are not audits, certifications, investment reports, purchasing advice, or evaluations of quality.
Onyx Security sells enterprise security teams software that finds approved and unapproved AI agents, monitors their prompts and actions, and enforces policies on them. Founded in 2024, it names Revolut among its customers, which it says include dozens of enterprises such as Fortune 500 banks and insurers. It has raised $153 million, including a $113 million Series B led by Bessemer Venture Partners. A Ctech article states that Onyx employs more than 80 people. Onyx’s durable advantages are those enterprise accounts and its Guardian Agent, a supervisory AI that can block an agent’s action or require human approval. Anthropic lists Palo Alto Networks and Okta beside Onyx in its security and compliance integration roster. Either could extend its products into the AI agent controls Onyx sells.
| Description | Onyx: Control plane that discovers sanctioned and shadow AI agents, monitors prompts and agent actions in real time, and enforces security and governance policies across enterprise AI use. | [f1] |
|---|---|---|
| Founded | 2024 | [f2] |
| HQ | Tel Aviv, Israel and New York, USA | [f3] |
| Funding | $153M total | [f4] |
| Latest funding | Series B, $113M (July 2026) | [f5] |
| Deployment | Hybrid, SaaS, Self-hosted | [f6] |
| Product | What it does |
|---|---|
| Onyx | Onyx: Control plane that discovers sanctioned and shadow AI agents, monitors prompts and agent actions in real time, and enforces security and governance policies across enterprise AI use. |
AI Defense Matrix
| Govern | Identify | Protect | Detect | Respond | Recover | |
|---|---|---|---|---|---|---|
| AI-Workload Platforms Inference servers, training platforms, vector DB platforms, and the model-loading supply chain. | ||||||
| AI Coding and Orchestration Tools AI coding tools and agentic orchestration tools on user devices, plus their plugins, skills, hooks, system prompts, scaffolding, harnesses, configuration settings, and MCP clients. | ||||||
| AI-Generated Code Code produced by AI tools, AI-assisted reviews, AI-generated infrastructure-as-code and tests, and vibe-coded apps that bypass CI/CD. | ||||||
| AI Traffic Paths to external and self-hosted AI services, MCP tool channels, agent-to-agent calls, model-registry downloads, and egress to unapproved AI services. AI gateways, LLM routers, and MCP gateways steer traffic along those paths. The steering decisions and their inputs belong here too: routing policies, MCP server registries, and agent naming and discovery services. | ||||||
| AI Model Model weights, fine-tuning checkpoints, model cards, registries, an AI bill of materials (AIBOM), and the third-party LLMs your enterprise consumes. | ||||||
| Training Data Datasets used for training, fine-tuning, and continued learning. | ||||||
| Runtime AI Data User prompts, inference inputs, RAG content, vector DB content, persistent agent memory, and interaction history. | ||||||
| AI Agent Identities AI agents as non-human principals, plus credentials, keys, permission scopes, service accounts, and delegation chains across agents and tools. |
Onyx is a control plane that discovers sanctioned and shadow AI agents, monitors prompts and agent actions in real time, brokers model-access traffic, and enforces and governs security policy across enterprise AI use. It is mapped to the AI Defense Matrix. [f7]
How well the company can compete in its security market, scored across eight dimensions against public evidence.
| Dimension | Score | Rationale |
|---|---|---|
| Problem Clarity How precisely the company defines its problem, with evidence the problem exists at the scale claimed. | 3/5 | Onyx names its buyer as the security, governance, and AI teams adopting agents. A SiliconANGLE article states that the problem is real as agents gain direct access to internal platforms. The cited press articles do not quantify that pain, so the problem is clear and credible without the independent quantification a higher score needs. [s1, s5, s3] |
| Capability Depth How specific the technical capabilities are, with evidence beyond marketing claims such as docs and third-party validation. | 3/5 | Onyx’s platform page documents concrete mechanisms, including an inline MCP gateway that logs every request, automated red teaming, session replay, and natural-language policies, and its Anthropic page describes ingestion through the Claude Compliance API. No open code, third-party evaluation, or customer technical writeup of the Guardian Agent appears in the reviewed sources, so the depth is documented by Onyx rather than validated outside it. [s10, s11, s2, s4] |
| Market Timing Whether the market is ready for this product, with evidence that buyers are actively seeking solutions. | 3/5 | The enabler is AI agents making real-time decisions inside enterprise systems, which a Ctech article states create a layer of risk existing security systems struggle to manage. Buyer-side demand includes Revolut’s documented use of Onyx, while investor rounds, Onyx’s own revenue claim, and Google’s 2026 cohort selection are indirect signals. [s3, s6, s8, s14] |
| Team Credibility Demonstrated domain expertise with public signals such as prior exits, publications, and industry recognition. | 3/5 | A Ctech article states that CEO Maxim Bar Kogan served in Unit 8200 and was Vice President of Product and Engineering at Mixtiles, and that co-founder Gil Elbaz is a veteran AI researcher who reported to the CTO of Nvidia. That is verifiable senior experience. The reviewed sources name no prior in-domain product build or exit for either founder, which a higher score requires. [s3, s6, s2] |
| GTM Proof Evidence of actual traction (customers, revenue signals, partnerships) beyond stated intentions. | 4/5 | Anthropic documents an Onyx integration in its Compliance API help center, with a setup guide, and lists Onyx among more than 60 security and compliance partner integrations. A Ctech article states that Anthropic announced it had integrated Onyx’s technology and that Onyx works with several Fortune 500 customers, and Onyx’s Series B release names Revolut through its Head of Security Engineering. That is verifiable partnership motion, multiply sourced. The fourfold revenue growth is Onyx’s own figure, so no third party reports scale. [s15, s16, s6, s8] |
| Funding Efficiency Whether funding matches go-to-market ambition, with signs of capital-efficient growth. | 2/5 | Onyx has raised $153 million in two years, including a $113 million Series B that a Ctech article states came four months after its $40 million launch, at an estimated $640 million valuation. No revenue figure is disclosed beyond Onyx’s own fourfold growth claim and the commercial proof is early, so the raise materially outruns verifiable commercial results despite visible shipping. [s7, s6, s8, s3] |
| Category Clarity Whether the company creates or fits a recognizable category that buyers can quickly place in their stack. | 3/5 | Press articles describe Onyx as an AI agent security company, and a Ctech article lists AI agent security and governance among six focus areas in Google’s 2026 startup cohort. Onyx’s own blog reports a Representative Vendor listing in Gartner’s February 2026 Market Guide for Guardian Agents, so the category is emerging and still new rather than independently established. [s4, s14, s9, s5] |
| Incumbent Defensibility How vulnerable the core value proposition is to absorption as a feature by a platform vendor. | 3/5 | Anthropic lists Palo Alto Networks and Okta beside Onyx in its security and compliance integration roster. A platform vendor would have to build the Guardian Agent and the proprietary models Onyx says track each step of an agent’s reasoning. The cited record does not show a moat that bundling could not eventually replicate. [s16, s15, s7, s2] |
Onyx Security sells control over the AI agents an enterprise runs, from coding assistants to agents embedded in SaaS tools. Its homepage names the buyer as the security, governance, and AI teams that need to protect, optimize, and measure agent adoption across the enterprise. A Ctech article states the underlying risk: agents make real-time decisions from external data and changing contexts, a layer existing security systems struggle to manage.
Independent coverage treats the problem as real. A SiliconANGLE article states that as organizations deploy agents across engineering, operations, and customer service, those systems gain direct access to internal platforms and business processes. Onyx argues, per the same report, that agents with this access expose companies to risks ranging from reasoning errors to prompt-based attacks that manipulate agent behavior.
The cited press articles do not quantify incidents or losses from agent misuse, so a buyer judging urgency has the adoption trend and Onyx’s own framing to go on. [s1, s3, s5]
The Onyx platform discovers the AI agents, models, and connected tools in an enterprise and turns governance policies into controls enforced at runtime, per the launch announcement. A SecurityWeek article states discovery across cloud, endpoints, code, and SaaS, with Onyx approving or correcting agent actions to enforce security and governance policies.
The Onyx Guardian Agent is the supervisory layer. The launch announcement says it can block an unsafe action, require human approval, narrow an agent’s scope, or redirect it before anything reaches a downstream system. A SecurityWeek article attributes to Onyx the claim that Onyx’s proprietary models track an agent’s decision-making at every step.
The platform page lists modules beyond supervision. They include an inline MCP gateway that logs every request and response, an AI gateway for model traffic, automated red teaming, session replay with an audit trail, and policies written in natural language. For Claude Enterprise, Onyx pulls conversation content, files, and activity logs through Anthropic’s Compliance API with no proxy in the path.
No open-source code, third-party evaluation, or customer technical writeup of the Guardian Agent appears in the reviewed sources, so a buyer would test the supervision claims in its own environment. [s2, s4, s7, s10, s11]
Onyx competes for the enterprise AI agent security budget against funded startups and against platform vendors that sell to the same security teams. Ctech lists Onyx among the companies Google selected for its AI cybersecurity startup cohort, where AI agent security and governance was one of six focus areas.
Anthropic lists Palo Alto Networks and Okta beside Onyx in its security and compliance integration roster. Onyx offers coverage across SaaS, cloud, endpoint, and code with more than 100 pre-built integrations, plus a Guardian Agent that can block or redirect agent actions.
A Bessemer post states that buyers chose Onyx after evaluating rival products in the category, citing fewer false positives. That account comes from the lead investor, which has a stake in Onyx, and names no customer. [s5, s14, s16, s10, s12]
Onyx’s customer evidence grew after its March 2026 launch. Its Series B release names Revolut as a customer, quoting Uros Solar, Head of Security Engineering, on using Onyx to govern coding agents across several service providers. Onyx also quotes two named CISOs, Eli Edelkind and Kyle Weckman, on its homepage, and the page text does not name their employers.
Most scale figures are Onyx’s own. The company says it has grown revenue fourfold since leaving stealth and is trusted by dozens of enterprises, including Fortune 500 companies in banking, insurance, and energy. A Ctech article states several Fortune 500 customers without naming them, and the homepage counts more than 1.1 million agents secured across deployments.
Ecosystem placement is documented outside Onyx. Anthropic lists Onyx among more than 60 security and compliance providers with Claude Compliance API integrations, and A Ctech article states that Google selected Onyx for its AI cybersecurity startup cohort. The motion is direct enterprise sales: Bar Kogan told Ctech at the Series A that Onyx had a US sales team, and the Series B funds sales, business development, and marketing in the United States and abroad. [s8, s1, s6, s15, s16, s14, s3]
Onyx’s founders pair a security background with AI engineering. A Ctech article states that CEO Maxim Bar Kogan served in Unit 8200 of the IDF and was Vice President of Product and Engineering at Mixtiles. Co-founder and Chief AI Officer Gil Elbaz is a veteran AI researcher who reported to the CTO of Nvidia, per Ctech, and the launch announcement says he led agentic AI work there.
The reviewed sources name no prior exit or security product build for either founder, so their credibility comes from their backgrounds. The team has grown quickly, from more than 70 people at the March launch to more than 80 by the July Series B, per SecurityWeek and Ctech. [s3, s6, s2, s4]
Onyx handles sensitive AI activity, so its own data handling is an early procurement question. Its Claude integration pulls conversation content, uploaded files, and activity logs into the Onyx control plane, and the platform page offers cloud, hybrid, or self-hosted deployment for data residency requirements.
Onyx’s trust center lists ISO/IEC 27001:2022, SOC 2 Type 2, and GDPR, and it offers an ISO 27001 report and a penetration-test report through an access request. It also lists a responsible-disclosure page and names Anthropic, OpenAI, and AWS among its subprocessors. [s11, s10, s13]
| Company | Relationship | Note | Compare |
|---|---|---|---|
| Pillar Security | competes with | Competes for the same enterprise security buyer governing AI agents and AI applications at runtime. | |
| Lasso Security | competes with | Competes with Onyx for enterprise budgets for monitoring and controlling generative AI and agent use. | |
| Capsule Security | competes with | Named beside Onyx in Ctech’s report on Google’s 2026 AI cybersecurity startup cohort, competing for buyers of runtime protection for AI agents. | |
| Bold Security | adjacent | Launched the same day as Onyx with the same $40 million raise, per SiliconANGLE, and focuses on protecting the endpoint layer. | N/AWe captured the evidence for these companies under different evidence-model versions (v1 vs v2), so the totals were scored under different conditions and are not directly comparable. |
| Palo Alto Networks | adjacent | Listed beside Onyx among Anthropic’s Compliance API partners and sells to the same enterprise security teams. | N/AWe scored these companies at different scopes, so the totals measure different things. |
| Okta | adjacent | Listed beside Onyx among Anthropic’s Compliance API partners, adjacent through identity and access governance for agents. |
Add analyzed competitors to compare them side by side with Onyx Security.
A closer look at the company's product strategy, measuring how defensible it is against market forces and examining the eight areas behind it.
reinforce or reposition
Onyx sells software that a customer’s security team operates, so its durable advantages are its engineering and its footing in large enterprise accounts. The Guardian Agent and the proprietary models Onyx says track each step of an agent’s reasoning are real-time machine learning work that would take a funded rival time to build. A customer that leaves rebuilds its integrations with discovery tools and rewrites its natural-language policies in another product. Its ISO 27001, SOC 2 Type 2, and GDPR listings appear in Onyx’s trust center without tying a buyer to Onyx. Anthropic lists Palo Alto Networks and Okta beside Onyx in its integration roster, so either could extend toward Onyx’s category.
| Dimension | Score | Rationale |
|---|---|---|
| Value Delivery Does the product sell software as the product, or judgment, trust, or accountability with software as the delivery mechanism. | 1/3 | Customers buy software, a control plane and Guardian Agent that their own security teams operate and configure with natural-language policies. The reviewed sources describe no managed service in which Onyx takes accountability for outcomes. |
| Switching Cost How expensive leaving is for a customer: data portability, integrations, learned workflows, network effects, regulatory data residency. | 2/3 | Onyx connects to browser, cloud security, network access, and endpoint detection sources. It ingests Claude activity through Anthropic’s Compliance API and can proxy Model Context Protocol traffic through an inline gateway. A customer that switches would need to rebuild those integrations and rewrite its natural-language policies. Switching mechanism documented. The cited record does not size the migration or show the inline gateway forcing a re-architecture on exit. |
| Compliance Moat Whether certifications, liability acceptance, or audit trails block an easy replacement. | 1/3 | Onyx’s trust center lists ISO/IEC 27001:2022, SOC 2 Type 2, and GDPR, attestations a funded competitor can obtain through ordinary enterprise preparation. No source shows a mandate or authorization that ties a buyer to Onyx. |
| Problem Complexity Whether the product requires ML, optimization, real-time systems, or years of specialized expertise. | 3/3 | Onyx says its proprietary models track an agent’s decision-making at every step and intervene as it happens, and its lead investor describes small purpose-trained models running in line on production traffic. That is machine learning and real-time systems work. |
| Buyer Profile Whether buyers are SMB operators, mid-market IT teams, or regulated enterprises and governments with procurement gates. | 3/3 | The evidenced buyers are large regulated enterprises. Onyx’s Series B release says its customers include Fortune 500 companies in banking, insurance, and energy and names Revolut through its Head of Security Engineering, and Ctech separately reports several Fortune 500 customers. The sector detail is Onyx’s own, while the Fortune 500 buyer class appears in independent press. |
| Layer Whether the product is an end-user application, a platform with application features, or infrastructure other applications depend on. | 2/3 | Onyx is a platform with application features, a console that observes and governs agents across SaaS, cloud, endpoint, and code. Its AI and MCP gateways run in line on the model and tool traffic that passes through them. Its Claude Enterprise integration uses Anthropic’s Compliance API with no proxy, and the cited record does not show applications depending on Onyx’s gateways to run. |
| Proprietary Data, Content, or IP Whether the product accumulates datasets, content licenses, or IP that a rival cannot recreate from scratch. | 1/3 | A SecurityWeek article and Onyx describe proprietary models, and a Bessemer post states that they learn how agents behave across the enterprise. No source names a retained dataset, states what the models accumulate from across customers, or shows an asset Onyx holds outside each tenant. Proprietary models named as a method do not evidence an accumulated data asset. |
Onyx sells to large enterprises whose security, governance, and AI teams govern agents across departments, per its homepage. The Series B release says its customers include Fortune 500 companies in banking, technology, insurance, and energy, and that Onyx controls both third-party agents and agents built by internal engineering teams. Ctech separately reports several Fortune 500 customers.
The Series B release names Revolut as a customer through its Head of Security Engineering. That buyer uses Onyx to govern coding agents across several service providers.
Onyx pitches to several teams inside an account. Its release describes a platform giving security, governance, and infrastructure teams unified control over agents, and the platform adds return-on-investment dashboards by department, so a single deal can involve buyers with different priorities.
The Onyx platform discovers AI assets across the enterprise, from agents and models to tools connected through the Model Context Protocol, and turns governance policies into controls enforced at runtime, per the launch announcement. A SecurityWeek article describes discovery across cloud, endpoints, code, and SaaS, with Onyx approving or correcting agent actions.
The Guardian Agent is the capability Onyx emphasizes. It can block an unsafe action, require human approval, narrow an agent’s scope, or redirect it before anything reaches a downstream system. SecurityWeek, attributing the claim to the company, reports that Onyx’s proprietary models track an agent’s decision-making at every step and intervene as it happens.
Bessemer, Onyx’s lead Series B investor, describes the model design: a mesh of small, purpose-trained models that run in line on production traffic without a noticeable latency cost. That description comes from an investor, and no benchmark or third-party evaluation of the models appears in the reviewed sources, so a buyer would verify accuracy and latency in its own trial.
Onyx runs a direct enterprise sales motion. A Ctech article states that Onyx had a sales team in the United States, and the Series B funds expansion of sales, business development, and marketing in the United States and other markets. The homepage routes prospects to a conversation with Onyx’s experts.
Customer evidence is partly named. The Series B release quotes Revolut’s Head of Security Engineering, and Onyx quotes two named CISOs on its homepage. Onyx says it serves dozens of enterprises, including Fortune 500 companies, and has grown revenue fourfold since launch, figures that are the company’s own.
Two ecosystem placements come from third parties. Anthropic lists Onyx among more than 60 Compliance API integration providers. A Ctech article states that Google selected Onyx for an AI cybersecurity startup cohort.
Onyx routes prospects to a demo request and a conversation with its experts. The page list in its sitemap includes a demo-request page and no pricing page, and the homepage routes visitors to talk with its experts, the negotiated-deal pattern of an enterprise sale.
The value Onyx describes scales with an enterprise’s agent footprint. It counts more than 1.1 million agents secured across deployments and offers return-on-investment dashboards that track adoption by department. The reviewed sources do not state whether Onyx charges by agent, by employee, or by platform, so a buyer learns the pricing unit in the sales process.
Onyx delivers a control plane that connects to the systems where agents already run. The platform page describes one console for agents and models across SaaS, cloud, endpoint, and code, fed by browser, AI platform, cloud security, network access, and endpoint detection sources, with more than 100 pre-built integrations.
Deployment options vary by customer. Onyx offers cloud, hybrid, or self-hosted deployment for data residency requirements. Its Claude coverage uses Anthropic’s Compliance API with no proxy, while its AI and MCP gateways run in line on the traffic they handle.
The reviewed pages describe capabilities and deployment modes but no service-level commitment or status page, so latency and failure behavior for the in-line components are questions a buyer settles with Onyx directly.
Onyx inspects sensitive AI activity, so a buyer reviews its own data handling early. Its Claude integration pulls conversation content, uploaded files, project data, and activity logs into the Onyx control plane, per its Anthropic integration page.
Onyx’s SafeBase trust center lists ISO/IEC 27001:2022, SOC 2 Type 2, and GDPR. It offers an ISO 27001 report and a penetration-test report through an access request, lists a responsible-disclosure page, and names Anthropic, OpenAI, and AWS among its subprocessors.
For its customers’ own compliance work, Onyx offers natural-language policies, session replay with an audit trail, and support for frameworks such as the EU AI Act. Those features help a buyer document AI governance, and Onyx releases the reports behind its own attestations through an access request.
Onyx positions its control plane above the AI platforms an enterprise already uses. Its homepage lists Anthropic, OpenAI, Amazon, Microsoft, and Google as integration partners, and the platform page counts more than 100 pre-built integrations across agent platforms, model providers, SIEMs, and clouds.
Anthropic confirms the Claude integration. A Ctech article states that Google selected Onyx for an AI cybersecurity startup cohort. Anthropic lists Onyx among its Claude Compliance API integration providers, and a Ctech article states that Google selected Onyx for its AI cybersecurity startup cohort.
The same channels are open to other vendors. Palo Alto Networks and Okta appear beside Onyx on Anthropic’s partner list, and Onyx’s Claude coverage depends on an interface Anthropic controls. Anthropic therefore decides what Claude data that interface gives Onyx.
Onyx’s founders pair a security background with AI engineering. A Ctech article states that Maxim Bar Kogan served in Unit 8200 of the IDF and was Vice President of Product and Engineering at Mixtiles. Co-founder Gil Elbaz is a veteran AI researcher who reported to the CTO of Nvidia, per Ctech. The launch announcement names Elbaz as Chief AI Officer and says he led agentic AI work with Nvidia’s CTO.
The reviewed sources name no prior exit or security product build for either founder, so the team’s credibility comes from its backgrounds. The company grew from more than 70 people at its March 2026 launch to more than 80 by the July Series B, per SecurityWeek and Ctech.
| Id | Source | Tier | Accessed |
|---|---|---|---|
| f1 | Onyx Security homepage | official | 2026-09-22 |
| f2 | CTech on the Onyx Security $35M Series A (March 12, 2026) | press | 2026-09-22 |
| f3 | SecurityWeek on the Onyx Security launch | press | 2026-09-22 |
| f4 | SecurityWeek on the Onyx Security $113M Series B (July 2026) | press | 2026-09-22 |
| f5 | Ctech on the Onyx Security $113M Series B (July 30, 2026) | press | 2026-09-22 |
| f6 | AI Defense Matrix Catalog entry | other | 2026-06-10 |
| f7 | AI Defense Matrix Catalog mapping | other | 2026-09-22 |
This site is an experimental research aid created by Zeltser Security Corp. All its data gathering and analysis was performed autonomously without human review, and it can contain errors of fact, interpretation, and judgment that a human reviewer might catch.
The analyses are statements of opinion, not statements of fact. Machine analysis produced the scores, summaries, and matrix placements by weighing the public sources each page cites, and reasonable people can weigh the same sources differently. Where a page states a fact, it cites the public source and the date it was checked, and the statement is only as accurate as that source. Unless a profile expressly says otherwise, the analysis involves no hands-on testing and no independent validation of any company's products or services.
Nothing here is professional, security, legal, financial, investment, or purchasing advice, and nothing here is a recommendation to invest in, do business with, or avoid any company. Inclusion of a company is not an endorsement, and absence of a company is not a judgment about it. Reading this site creates no advisory or client relationship. Verify any detail you plan to act on against the vendor's current materials.
The content is provided "as is" and "as available," with all warranties disclaimed, express or implied, including merchantability, fitness for a particular purpose, accuracy, and non-infringement. No entry is warranted to be complete, current, or correct. Companies change, vendors update their claims, sources can be wrong, and automated analysis can misread them.
To the fullest extent permitted by law, the operator, Zeltser Security Corp, is not liable for any damages that arise from using this site or relying on its content, including direct, indirect, incidental, special, and consequential damages and lost profits, even if advised that such damages were possible. If you are dissatisfied with the site or disagree with these terms, your remedy is to stop using it.
Entries link to vendor pages, press coverage, and other external sites that Zeltser Security Corp does not control and is not responsible for. A link is not an affiliation with the destination or an endorsement of it. Product and company names and trademarks are the property of their owners, used here nominatively to identify the companies described. Short quotations from cited sources appear for identification and commentary.
Use, quotation, automated retrieval, and redistribution of the content are governed by the Terms of Use at cybercompanyprofiles.com/terms, which permit personal and internal business use with attribution and prohibit republication and resale.