Copperhelm

• Wiz, Palo Alto Networks, or CrowdStrike could attach autonomous investigation and remediation agents to the cloud asset graphs they already maintain, turning Copperhelm’s pitch into a feature of a product the buyer already licenses.
• Enterprise change-control teams could confine Copperhelm’s agents to read-only recommendations, which would erase the difference between Copperhelm and the prioritization features cloud security incumbents already ship.
• A faulty automated remediation, such as a WAF rule that blocks production traffic, would undercut the safety record that selling autonomous remediation to regulated buyers depends on. This risk is hypothetical, and no such failure appears in the public record.
• Copperhelm’s agents reason with frontier models such as Claude, so a model provider’s pricing or safety-policy change could raise Copperhelm’s costs or degrade its capability without any competitor making a move.
• Copperhelm publishes one marketplace price, a $50,000 twelve-month Protect workloads contract, but no detailed packaging, usage metrics, or documentation, so buyers who cannot evaluate the product without a sales conversation may default to suites they already license.

Copperhelm sells automation for cloud security work that stayed manual while engineering work did not. Its about page argues that software engineers adopted AI years ago while cloud security teams still sift millions of raw alerts by hand, and the company offers autonomous agents that investigate, validate, and fix issues rather than report them. The buyer is a cloud security or vulnerability management team at a large enterprise running infrastructure across hundreds of cloud accounts.

The vendor quantifies the pain through its launch case study. The press release describes a Fortune 500 company whose 6 million raw findings became a shortlist of a few hundred validated risks, a vendor claim with no independent measurement behind it. The Mythos page adds urgency with a vendor-authored scenario in which AI attackers collapse the exploit window to 15 minutes, making 14-day patch cycles obsolete.

Independent corroboration of the problem stops at relay. SiliconANGLE and Ynetnews repeat the company’s framing that cloud environments are too fragmented for manual work or generic AI, and the buyer-side voices that confirm the pain, security leaders at Fiserv, Gong, and Cass Information Systems, speak in testimonials on the vendor’s own site. [s2, s8, s4, s10, s11, s3]

Copperhelm documents four named agents with mechanisms specific enough to evaluate. The System Agent connects to running machines and inspects active processes and in-memory code to judge whether a vulnerability is actually loaded. The Network Agent maps cloud topology and traces reachability through VPCs, security groups, and WAFs. The Adversary Agent simulates external attack paths with a benign payload to prove exploitability, and the Resolution Agent deploys WAF rules and generates context-rich prompts for AI coding assistants, a roster the homepage names as Cursor, Claude Code, OpenAI Codex, GitHub Copilot, Antigravity, and Grok Build.

The Context Lake is the claimed foundation under the agents. The company describes it as a real-time ontology of the customer’s multi-account cloud architecture, and the Fiserv testimonial states the problem it solves, that an LLM dropped into 200 AWS accounts cannot know how anything connects. A Build Your Agent option invites customers to connect their own AI workflows to the same data layer.

Verification stops at the marketing site. Copperhelm publishes no documentation portal, no self-service demo, and no third-party technical evaluation, and the AWS Marketplace listing repeats the vendor’s own description. The company announced joining the Anthropic Cyber Verification Program, an access program for security firms using frontier models, which signals vetting for model access rather than independent validation of the product. [s3, s12, s5, s1]

Copperhelm positions against scanners and manual triage rather than against named vendors. The AWS Marketplace listing claims the product does what an engineer would do without the engineer, validating exploitability at runtime instead of guessing from configuration, and the about page claims Copperhelm invented the agentic cloud security category, a vendor superlative no third party has tested.

The crowded ground sits one category over. Cloud security platforms such as Wiz and Orca Security already sell exposure prioritization built on cloud asset graphs, Tamnoon sells agentic cloud remediation with human experts in the loop, and AI SOC vendors such as Torq and Dropzone AI sell autonomous investigation to the same security operations budget. Copperhelm’s wager is that validated exploitability plus autonomous remediation is a product of its own rather than the next feature of those neighbors.

The differentiation claim is depth at runtime. Checking whether a vulnerable package is loaded in memory and proving exploitability with a benign payload go beyond configuration scanning, but the incumbents already hold the environment data such checks need, so the visible edge is execution speed rather than structural advantage. [s12, s2, s3, s10]

Copperhelm runs a demo-gated, founder-fronted enterprise motion. The website’s only call to action is a demo booking, the AWS Marketplace listing exposes a $50,000 twelve-month Protect workloads contract while broader terms stay negotiated, no self-service trial appears in the fetched record, and CEO Shimon Tolts fronts the public storytelling in the launch press. The careers page as of July 2026 lists two openings, a Full Stack Engineer and a Senior Full Stack Developer, both in Tel Aviv, so no go-to-market organization is visibly being hired, which fits a seed-stage company whose founders still carry sales.

Traction claims arrived with the launch and have not yet been independently tested. The press release states the company already works with paying customers including Fortune 500 enterprises, independent press relays the same claim, and named practitioners at Fiserv, Gong, and Cass Information Systems describe deployments in site testimonials. No case study with measured outcomes has been published beyond the 6-million-findings figure in the vendor’s own release.

Distribution is one channel deep. The AWS Marketplace listing lets enterprises buy through committed cloud spend, and no reseller, MSSP, or technology-partner program appears in public materials. The Anthropic Cyber Verification Program announcement doubles as a credibility channel aimed at buyers nervous about autonomous AI. [s1, s6, s8, s9, s3, s12, s5, s10]

Copperhelm’s three founders carry verifiable cloud pedigree and relayed recognitions. Shimon Tolts is CEO, Roman Labunsky is CTO, and Eyar Zilberman is CPO, and independent press confirms prior leadership and senior technical roles at Unity, McAfee, and RSA. Ynetnews relays that the team includes people recognized as AWS Heroes, a CNCF Ambassador, and a GitHub Star, community standing in cloud infrastructure rather than a security research record.

The board pairs Israeli venture experience with a U.S. go-to-market investor. Rona Segev and Brian Sack of TLV Partners and Shay Michel of Merlin Ventures sit on the board per the company’s about page, and Segev’s launch quote vouches for the founders as cloud veterans. Merlin Ventures’ involvement is a board seat in the public record, and Copperhelm does not yet appear on Merlin’s published portfolio page.

Public gaps remain normal for the stage. No founder exit, security publication record, or conference footprint appears in the fetched sources, and the recognitions are relayed by press rather than confirmed against the awarding organizations. [s2, s9, s11, s8]

Copperhelm finished its trust homework before it announced itself. The trust center lists a SOC 2 Type II report covering 2025, audit work that ran while the company was still in stealth, alongside penetration testing, documented controls across product security, access management, and continuity, and a subprocessor list naming AWS, GitHub, and Google Workspace.

The product’s access model is the trust question the vendor’s own pages answer in two directions. The homepage deployment section asserts "Read-Only Access: Zero risk to live operations," while the same page states that real threats "get mitigated automatically using your existing WAF, EDR and Cloud settings" and the technology page describes a Resolution Agent that deploys WAF rules to mitigate active threats. Both statements are vendor claims, the reviewed pages do not reconcile how automated WAF changes operate under read-only access, and no published threat model, rollback discipline, or failure-mode description addresses what happens when an agent acts wrongly.

The Anthropic Cyber Verification Program announcement targets exactly this anxiety. The vendor frames membership as rigorous testing and validation of how its agents use frontier models, and the program itself is an application-based access pathway for legitimate security work, a meaningful signal of intent rather than a certification. [s7, s1, s3, s5]

Add analyzed competitors to compare them side by side with Copperhelm.

Copperhelm targets large enterprises whose cloud estates are too big for manual security work. Ynetnews describes a platform designed for cloud infrastructure that operates across hundreds of accounts, the launch release claims paying customers including Fortune 500 enterprises, and the Fiserv testimonial describes an environment of 200 AWS accounts. Regulated buyers are prominent by design, with Cass Information Systems introduced in its testimonial as a century-old, highly regulated financial institution.

The personas Copperhelm courts are named in its own testimonials. A director of vulnerability management at Fiserv, a director of product security at Gong, a CISO at Cass speaks for a deployment, and a CISO at JB Poindexter endorses the approach, which places the product in vulnerability management and cloud security operations budgets rather than a new line item. A former ServiceNow CISO endorses the approach without claiming a deployment.

The demand evidence is launch-quarter fresh, which cuts both ways. The company exited stealth on April 23, 2026, so the testimonials and the Fortune 500 claim are current by definition, and the reviewed public sources show no demand proof beyond the launch push itself apart from the Anthropic program announcement on the company’s site, which carries no publication date.

Copperhelm’s capability claims are specific enough to falsify, which is rare at this stage. The System Agent inspects active processes and in-memory code to judge whether a vulnerable package is actually loaded, the Network Agent traces reachability through VPCs, security groups, and WAFs, the Adversary Agent proves exploitability with a benign payload, and the Resolution Agent deploys WAF rules and writes context-rich prompts for AI coding assistants, a roster the vendor's pages name as Cursor, Claude Code, OpenAI Codex, GitHub Copilot, Antigravity, and Grok Build. That last detail concedes something honest, that permanent code fixes belong to the customer’s engineers and Copperhelm’s job is to justify the priority.

The claimed data advantage is architecture, not exclusivity. The Context Lake structures a real-time ontology of the customer’s multi-account cloud, and the Fiserv testimonial argues a bare LLM cannot know how 200 AWS accounts connect. The data underneath comes from the customer’s own environment, so the advantage is freshness and structure that a competitor with equivalent access could rebuild.

The AI description names real dependencies rather than vague intelligence. The Anthropic partner page states the agents use frontier models such as Claude for network analysis, live remediation, and vulnerability management, with the Cyber Verification Program participation described as testing and safety collaboration rather than a benchmark. No public documentation, demo, or third-party evaluation lets a buyer test any of these claims before a sales conversation.

Copperhelm sells founder-first with a demo emphasis, which fits its stage. The vendor site emphasizes demo booking beside a See Pricing path, AWS Marketplace adds purchase options and proposals, CEO Shimon Tolts carries the public narrative in launch press, and the careers page as of July 2026 lists two openings, a Full Stack Engineer and a Senior Full Stack Developer, both in Tel Aviv, so no quota-carrying sales organization is visibly forming ahead of repeatable proof.

Distribution is one channel and one credibility program deep. The AWS Marketplace listing provides an enterprise procurement channel with contract purchase options, no reseller or MSSP program appears publicly, and the Anthropic Cyber Verification Program announcement works as a trust signal aimed at buyers wary of autonomous AI. The board seat held by Shay Michel of Merlin Ventures, a firm that markets itself as a go-to-market investor with U.S. government reach, hints at a federal channel ambition the public record does not yet show.

Customer proof is testimonial-grade rather than case-study-grade. Press repeats the vendor's customer and adopter claims while only the vendor release describes them as paying, named practitioners vouch on the vendor site, and the single quantified outcome, 6 million findings reduced to a few hundred validated risks, appears in the vendor’s own release without measurement methodology.

Copperhelm publishes one concrete price, and only on its marketplace channel. The AWS Marketplace listing posts a 12-month contract at $50,000 for a dimension it labels Protect workloads, and the listing states that pricing is based on the duration and terms of the contract with the vendor. The 24-month and 36-month tabs show no concrete figures in the reviewed listing, so that single published anchor, paired with the demo gate, signals negotiated enterprise deals rather than self-service or metered adoption, and the reviewed pages otherwise show no list price.

The charging unit is partly disclosed and mostly open. The marketplace dimension names workloads, so workload-based packaging is publicly indicated, but the reviewed listing does not disclose how many workloads the $50,000 tier covers, what overage costs, or whether agent activity is metered separately. For a product whose agents consume frontier-model compute at investigation time, the gap between a fixed-tier price and variable cost is a real margin risk the public materials never address. The twenty-senior-engineers framing in the CEO’s launch quote implies pricing against replaced labor, the unit a security leader would use to justify the spend, but no published evidence confirms an account-based, activity-based, or flat-platform model beyond the workload dimension.

Copperhelm describes its deployment model on the homepage at claim level rather than document level. A deployment section promising a 14-day rollout states "100% Sensorless: No heavy software to install," "Read-Only Access: Zero risk to live operations," and "Cloud-Native APIs: Seamless, instant connection." Those lines are vendor claims without published permission scopes, architecture references, or install documentation behind them, so a buyer still cannot verify the operational footprint before a sales conversation.

The vendor’s access claims pull in two directions. The same homepage that asserts "Read-Only Access: Zero risk to live operations" also states that real threats "get mitigated automatically using your existing WAF, EDR and Cloud settings," and the technology page describes a Resolution Agent that deploys WAF rules to mitigate active threats. The reviewed pages do not reconcile how automated mitigation operates under read-only access: code fixes are described as ready to approve and deploy, but no published rollback path or blast-radius limit describes what happens when an automated infrastructure change goes wrong. For the regulated enterprises Copperhelm courts, that unreconciled change-management story is the deployment decision.

Hosting signals are conventional. The trust center lists AWS as the cloud service provider and GitHub and Google Workspace as subprocessors, and no contractual uptime, support-response, or service-level commitment appears in the reviewed materials, distinct from the 15-minute remediation-speed framing the product marketing uses.

Copperhelm leads with trust collateral. The trust center lists a SOC 2 Type II report labeled 2025, with no issuance date shown in the reviewed materials, alongside penetration testing, documented controls across product security, access management, data security, and continuity, and a subprocessor list. A populated trust center is a posture aimed at the regulated buyers Copperhelm courts.

The Anthropic Cyber Verification Program membership extends the same strategy to AI anxiety. The vendor describes rigorously testing and validating how its agents use frontier models, so the membership signals a program relationship the vendor itself describes rather than third-party validation of the product.

What remains unaddressed is the failure story. The trust center confirms an incident response plan was tested as a SOC 2 control, but no public vulnerability disclosure policy, agent-error threat model, or customer-facing incident-response and SLA commitment appears in the reviewed materials. For a vendor that sells automated mitigation of live threats, the absent answer to what happens when the agent is wrong is the central open trust question.

Copperhelm ships an application and markets a platform. The product its agents deliver, validated exposures and executed fixes, is a tool for one security team, while the Build Your Agent pitch invites customers to connect their own AI workflows to the Context Lake and build agentic solutions on what the vendor calls its architectural data foundation. No public API documentation, partner program, or third-party integration substantiates the platform half yet.

The integration story runs toward engineering workflows rather than the security stack. The Resolution Agent hands remediation context to AI coding assistants such as Cursor and Claude Code, a forward-looking choice that meets developers where fixes actually happen, while conventional security integrations, SIEM, SOAR, and ticketing, go unmentioned in public materials.

Ecosystem positioning today reduces to two dependencies and one listing. Copperhelm builds on Anthropic’s models, hosts on AWS, and sells through AWS Marketplace, so its ecosystem leverage is borrowed from larger platforms rather than owned, the normal posture for a seed-stage company and a dependent one for an autonomy product built on another vendor's frontier models.

Copperhelm’s founding team is cloud-native in a category where most founders are security-native. Press verifies that CEO Shimon Tolts, CTO Roman Labunsky, and CPO Eyar Zilberman held leadership and senior technical roles at Unity, McAfee, and RSA, and Ynetnews relays team recognitions as AWS Heroes, a CNCF Ambassador, and a GitHub Star. The pedigree matches the product thesis, that cloud architectural understanding is the scarce ingredient in cloud security automation, and TLV Partners’ Rona Segev publicly vouched for the founders as cloud veterans.

The board concentrates venture and channel experience around a three-person founding team. Rona Segev and Brian Sack of TLV Partners and Shay Michel of Merlin Ventures hold the named board seats, and Merlin’s self-described go-to-market and U.S. government focus complements an Israeli company that must eventually sell into American regulated enterprises. The board seat is the verifiable extent of that relationship in the reviewed record.

Execution gaps are the stage-normal ones. The visible team is small, hiring as of July 2026 is limited to two full stack engineering roles in Tel Aviv, the reviewed sources identify no security research published under the company’s name, and the recognitions that anchor team credibility are relayed by press rather than independently confirmed.