Token Security

• Identity incumbents could bundle agent-identity governance within a few years and close the standalone window. Cisco is folding Astrix Security into its identity portfolio, and CyberArk and other identity incumbents sell adjacent suites to the same buyers.
• The triple-digit growth claim is vendor-issued with no disclosed revenue or customer counts, so actual traction could be thinner than the logo wall and the eight Gartner reviews suggest.
• A non-human identity rival such as Oasis Security or Entro Security could land deeper identity-provider partnerships and relegate Token Security to a niche.
• Emerging agent-identity standards could let cloud and identity providers issue and govern agent credentials natively, which would shrink the discovery and governance gap Token Security monetizes.
• An enterprise pullback in agentic-AI deployment would defer the AI-agent identity budget line and return the company to the slower service-account hygiene sale.

Token Security targets the gap between how enterprises govern human identities and how they govern machine ones. The company says AI agents and non-human identities span on-prem, cloud, and AI platforms while outnumbering human identities by 50 to 1. Service accounts, API tokens, and autonomous agents hold credentials that rarely pass through joiner-mover-leaver review, so over-privileged identities persist.

Independent reporting corroborates the pain without confirming the vendor's ratio. TechCrunch tied machine-identity risk to outdated or over-permissioned credentials and quoted CEO Itamar Apelblat saying hackers log in rather than break in, and Calcalist framed the same problem as the reason enterprises are funding the category. The buyers Token names, the security, IAM, and cloud teams, are the teams that carry the remediation work.

Agentic AI shifts the problem from hygiene to runtime control. Agents act autonomously at machine speed, and a single compromised agent can reach every system it connects to, which moves the buyer's question from inventory to enforcement. [s1, s9, s10, s5]

The Token Security platform covers the machine-identity lifecycle from discovery to remediation. Documented modules include continuous discovery and visibility, lifecycle management, security posture management, identity threat detection and response, and AI-driven automation and remediation, and the platform applies intent-based permissioning so an agent keeps only the access its purpose requires. The discovery module inventories AI agents and MCP servers alongside conventional service accounts and keys.

Token ships AI-native interfaces on top of the platform. The Token MCP Server and AI Agent give security teams a natural-language way to query inventory and posture from tools such as Claude or Cursor, a June 2026 Entitlements Inventory assigns each entitlement a risk level so teams triage the riskiest access first, and a June 2026 integration brings Anthropic's Claude Compliance API into the same governance model so Claude keys, workspaces, and agents are discovered and governed with other non-human identities.

The verifiable footprint is solid for the category, but external validation is thin. There is no public docs portal or third-party technical evaluation, the open-source footprint is a small Custom-GPT discovery tool rather than an open platform, and the eight reviews on Gartner Peer Insights speak to customer satisfaction rather than to capability depth. The inventory and risk map are built per customer, so the engineering is the differentiator rather than any data asset. [s1, s2, s3, s7, s8, s12]

Token Security competes in a non-human identity field that platform vendors are now buying into. Oasis Security and Entro Security are named non-human identity peers, and Gartner groups Token Security with Entro and Akeyless in a Workload Identity Management market. The incumbents arrived through acquisition, with CyberArk buying Venafi for machine-identity management and Cisco moving to acquire Astrix Security for a reported $400 million.

Token positions on AI-agent identity rather than the broader machine-identity catalog alone. The about page frames the company as the identity layer that makes agentic AI possible, and the platform leads with creation-to-runtime control of AI agents to reach the buyer feeling autonomous-agent risk now. That framing trades part of the broader machine-identity market for the faster-moving agent slice.

The acquirers that validate the market also become the competition. Cisco plans to fold Astrix into its identity portfolio, and CyberArk and other identity incumbents sell identity suites to the same buyers, so Token needs depth a bundled feature cannot match. [s12, s15, s11, s4, s2]

Token Security backs its traction with named enterprise customers. Calcalist and TechCrunch name HPE and HiBob in independent reporting, the 2025 growth release adds Bloomreach, BetterHelp, and Dayforce, and the trust center lists reviewers including Dayforce, AlphaSense, and ZoomInfo.

Third-party signals reinforce the logo wall. RSAC named Token Security a 2026 Innovation Sandbox finalist, and Gartner Peer Insights shows a 4.7 average across eight customer reviews in the Workload Identity Management market. The company claims triple-digit growth for 2025 in its own release, though revenue and customer counts stay undisclosed.

A hired go-to-market team now fronts part of the selling. The 2025 growth release is datelined New York, which signals a United States enterprise motion layered on top of the Tel Aviv engineering base. [s10, s9, s5, s6, s4, s14, s12]

Token Security's credibility comes from a founding team with verifiable Unit 8200 depth. TechCrunch reports that CEO Itamar Apelblat worked on defensive security at Israel's Unit 8200 while CTO Ido Shlomo led offensive work against nation-state targets, and the company is registered as a private limited entity, Token Security Ltd, in Tel Aviv. Public sources show no prior company built and exited by either founder and no sustained publication record.

Veteran investors back the team where the public track record is thin. Calcalist reports a $20 million Series A led by Notable Capital, with participation from TLV Partners and executives from Palo Alto Networks, CrowdStrike, Check Point, and Venafi. Those checks vouch for the founders ahead of disclosed revenue. [s9, s13, s10]

Token Security publishes the trust collateral enterprise buyers expect before granting broad identity access. The SafeBase-hosted trust center lists ISO/IEC 27001:2022 and SOC 2 Type 2 certifications, a pentest report and network diagram behind access gates, and named reviewers including HiBob, PROS, Dayforce, AlphaSense, BetterHelp, and ZoomInfo.

The collateral matters because the product needs privileged reach to do its job. Token discovers and governs identities across cloud, SaaS, CI/CD, source control, and AI platforms, so a buyer is granting visibility into its credential estate, and audit evidence is the precondition for that grant. The certifications ease procurement but do not block a substitute, since every same-stage rival can obtain the same audits and no regulation mandates this product class. [s6, s8]

Add analyzed competitors to compare them side by side with Token Security.

Token Security sells to enterprises whose machine identities have outgrown the controls built for people. The company frames service accounts, API tokens, and autonomous agents as the population legacy joiner-mover-leaver governance never covered, and it names security, identity and access management, and cloud teams as the buyers who carry that work. The about page positions the company as the identity layer that makes agentic AI possible.

Token now leads with agentic-AI messaging while still serving the broader non-human identity estate. It secures every non-human identity from service accounts to AI agents, but the messaging leads with creation-to-runtime control of agents, which targets the buyer feeling autonomous-agent risk now rather than the buyer doing routine service-account hygiene. That emphasis may concentrate demand among buyers with urgent agent-governance needs rather than the broader machine-identity base.

Named demand spans recognizable enterprises rather than design partners alone. Calcalist reports HPE and HiBob as customers, and the trust center names HiBob, PROS, Dayforce, AlphaSense, BetterHelp, and ZoomInfo among its reviewers. The open question is whether the agent-identity pitch pulls a budget line distinct from service-account spend or sells into the identity budget rivals already contest.

Token Security covers the machine-identity lifecycle from discovery through automated remediation. The platform runs continuous discovery and visibility, lifecycle management, security posture management, identity threat detection and response, and automation and remediation, and the product page describes intent-based permissioning that aligns an agent's access with what it is meant to do rather than the credentials it holds. That intent model replaces long-lived, over-scoped credentials with purpose-bound access and right-sizes permissions as agent behavior evolves.

The AI-native interfaces sit on top of that platform rather than replacing it. The Token MCP Server and AI Agent let teams query inventory and posture in natural language from tools such as Claude or Cursor, a June 2026 Entitlements Inventory ranks every entitlement by risk so teams triage the riskiest access first, and a June 2026 integration brings Anthropic's Claude Compliance API into the same model so Claude keys, workspaces, and agents are governed alongside other non-human identities. The cadence shows a team extending coverage toward the AI platforms its buyers adopt.

The public footprint is credible, and it is engineering depth rather than a data moat. The inventory and risk map are built per customer from the identity logs and telemetry Token says it analyzes, and the natural-language layer is an interface over the same platform data, reachable from common chat applications. A metadata flywheel across customers is conceivable for an inventory product of this kind, but the cited sources identify no proprietary cross-customer dataset or exclusive data right.

Token Security runs a demo-led enterprise motion with a public proof-of-demand record for its stage. Every product and platform page routes to a book-a-demo call rather than self-serve signup, which signals a sales-assisted enterprise purchase. The company reached named enterprises within months of leaving stealth, and Calcalist reported HPE and HiBob as customers eight months after it emerged from stealth.

External validation arrived faster than disclosed revenue. RSAC named Token a 2026 Innovation Sandbox finalist, Gartner Peer Insights lists it in the Workload Identity Management market with a 4.7 average across eight customer reviews, and the company claims triple-digit 2025 growth in its own release. Peer Insights ratings are end-user opinions rather than analyst-authored placement, so these are program and buyer-side signals on top of named-customer adoption.

Investor and acquirer behavior frames the commercial context. TechCrunch reported a $20 million Series A bringing total funding to $27 million, and the same identity vendors that validate the category are buying into it, with Cisco moving to acquire Astrix and CyberArk completing its Venafi purchase. Disclosed revenue or a customer count beyond the named logos does not appear in fetched sources.

Token Security publishes no pricing in fetched sources, so the charged unit and list price stay private. Every conversion path on the product and platform pages is a book-a-demo call rather than a published plan, which usually signals negotiated enterprise deals. The absence withholds any budget-anchoring signal, so a buyer cannot size the spend before entering a sales conversation.

The charged unit is not stated, though the product shape suggests what buyers pay for. Token meters a population it discovers, the non-human identities and agents across cloud, SaaS, CI/CD, source control, and AI platforms, so the natural unit is identities or agents under management rather than human seats. Confirming whether the meter is identities, integrations, or a platform subscription would require a sales conversation.

The model reads as capacity for governing a growing identity estate rather than a per-feature license. The agent-first framing and the intent-based control plane point at a buyer paying to bring an expanding machine-identity population under one policy, but the specific consumption terms stay behind the demo gate.

Token Security delivers as software the customer configures and operates, not as a managed service. The platform discovers, inventories, and governs non-human identities across on-prem, hybrid, and cloud environments, and the buyer runs it against its own estate rather than handing the function to Token analysts. Nothing in fetched sources describes a managed-detection or analyst-staffed operating model alongside the platform.

The AI-native layer changes how operators interact with the platform rather than who runs it. The Token MCP Server and AI Agent let teams query inventory and posture and request guided remediation in natural language, and automation and remediation modules act on findings the platform surfaces. Operators meet the platform inside the tools they already use.

Operational maturity collateral is partial. The trust center lists a network diagram and a penetration-test report in its document library, with sensitive information available through a request-access flow, which is more operational evidence than many same-stage rivals show, but published uptime commitments or support service-level agreements do not surface in fetched pages.

Token Security publishes the trust collateral an enterprise buyer expects before granting broad identity access. The SafeBase-hosted trust center lists ISO/IEC 27001:2022 and SOC 2 Type 2 certifications and a documents library that lists a pentest report and a network diagram, with sensitive information available through a request-access flow. The certifications and request-gated reports are stronger procurement evidence than a footer badge alone.

The collateral matters because the product needs privileged reach to do its job. Token discovers and governs identities across cloud, SaaS, CI/CD, source control, and AI platforms, so a buyer is granting visibility into its credential estate, and audit evidence is the precondition for that grant. The trust center displays HiBob, Dayforce, and ZoomInfo under the heading "reviewed and trusted by", which is the vendor's own framing rather than published evidence that each completed the review.

The attestations are enterprise-grade but table stakes rather than a moat. The two certifications ease procurement, and neither one establishes a Token-specific mandate, retained liability, or an authorization a replacement would have to reproduce, so they do not block a substitute. A buyer should still resolve data-handling and retention terms in a formal review beyond the published certifications.

Token Security positions as the identity control plane for non-human and agent identities rather than a point scanner. The Claude integration blog calls the product a unified control plane that brings AI-platform identities into the same governance model as other non-human identities, and the about page frames the company as the identity layer that makes agentic AI possible. The platform claim rests on becoming the place agent access is defined and enforced.

That control-plane position is a sharper bet than discovery alone, because intent-based least-privilege defines and enforces what agents may do. The Claude blog casts agent workflows as identity-driven access paths and Token as the control plane that secures them, so the product could become access-path infrastructure if customers adopt it for enforcement rather than only inventory. How deep that runs turns on whether buyers route agent governance through Token or query it occasionally.

The exposure is that adjacent identity vendors already own the control plane for humans. The incumbents arrived in non-human identity through acquisition, with Cisco moving to buy Astrix and CyberArk completing its Venafi purchase, so the agent-identity layer Token is building is ground established identity platforms are positioned to extend into.

Token Security's credibility comes from a founding team with verifiable national-security depth. TechCrunch reports the company was founded in 2023 and ties the founders to Israeli signals-intelligence backgrounds, with Apelblat on defense and Shlomo on offensive nation-state work. The cited sources describe entrepreneurial ventures and startup experience without identifying a prior company the founders built and exited.

The investor bench reinforces the founder signal where the public track record is thin. Calcalist reports the $20 million Series A was led by Notable Capital with total funding reaching $27 million, drawing participation from TLV Partners and executives from Palo Alto Networks, CrowdStrike, Check Point, and Venafi. That backing across an early round is itself a signal about the founders.

The depth below the founders is the open question. Fetched sources establish the founding team and the lead investor but do not surface individual prior builds across the broader leadership bench, so the verifiable strength is the founders' domain background and the investor confidence it has drawn.