Kong Inc.

• Cloud and model platforms that already sit in front of AI traffic, such as the major cloud providers, could bundle prompt screening and traffic governance into their own gateways, which would let buyers govern AI traffic without adding Kong's AI Gateway.
• The public record reports revenue and customer counts for all of Kong, not the AI Gateway, so the AI line's actual adoption is unproven and could be far smaller than the 700-customer installed base implies.
• Specialist AI-gateway vendors such as Portkey, BerriAI, and NeuralTrust compete directly on guardrails and multi-LLM routing, and a buyer choosing AI governance as a separate purchase from its API gateway could pick one of them over Kong.
• Kong's pivot to position itself as an AI connectivity company depends on enterprises treating the same control point for API, LLM, and agent traffic, and buyers who split those purchases would weaken the installed-base advantage the AI Gateway relies on.
• Most of the AI Gateway's capability evidence is Kong's own documentation and launch announcements, so an independent benchmark or a customer reporting weak guardrail accuracy could undercut the depth the product claims.

Kong sells the AI Gateway to enterprises that are moving AI from pilots into production and need to control the traffic. The company frames the buyer as engineering and platform teams who lack full visibility into AI usage and resource consumption, struggle to control token spend, and cannot maintain audit trails across agent workflows. The product sits between applications and model providers so that access, data leakage, and token usage become governed in one place.

Press coverage corroborates that the pain exists at enterprise scale. Reporting on the Agent Gateway launch described businesses moving AI systems from pilot projects into production and facing growing pressure to monitor usage, control costs, and maintain audit trails across increasingly complex workflows. That framing comes from trade press rather than Kong's marketing, which is stronger evidence that buyers are searching for the capability.

The asset under management is the AI request and the agent interaction. Kong governs how developers, apps, and agents consume LLMs, and the more recent Agent Gateway extends that to traffic between agents, so the boundary the product addresses is the point where prompts, responses, and agent actions meet privileged model access. [s1, s5]

The Kong AI Gateway is a governance layer built on top of Kong's existing API gateway, with named controls rather than marketing generalities. The documentation describes a semantic prompt guard that builds allow and deny lists of topics across every LLM, and a PII sanitization plugin that detects and redacts sensitive data across 20 categories and several languages and most major AI providers. A public documentation portal lets a buyer inspect how each control works.

Kong has extended the product from model traffic to agent traffic over successive releases. The April 2025 release focused on LLM-level governance, including the PII sanitization plugin and automated retrieval pipelines aimed at reducing hallucinations, and the April 2026 Agent Gateway release added governance of agent-to-agent communication alongside LLM and Model Context Protocol traffic. That progression tracks the market's shift from single model requests toward multi-step agent interactions.

What the public record does not yet include is independent validation of how well these controls perform. The capability is documented and specific, but the evidence is Kong's own documentation and launch announcements rather than a third-party benchmark or a customer technical writeup, which is the checkpoint that would lift the depth further. [s2, s4, s5]

Kong competes for the AI-governance buyer against specialist AI-gateway startups and the platforms that already carry AI traffic. Pure-play vendors such as Portkey, BerriAI, and NeuralTrust pair a gateway with guardrails and multi-LLM routing, overlapping Kong's runtime governance. On the other side, cloud and model platforms that sit in front of AI traffic can bundle the same screening and routing into their own offerings.

Kong's differentiator is the installed base its rivals lack, since the AI Gateway layers onto the API gateway that more than 700 enterprises already run. For a company already operating Kong, adding AI governance is a plugin on infrastructure it already maintains rather than a new vendor decision, which is a distribution advantage a standalone tool cannot easily copy. That same advantage is what the company is betting its repositioning on.

The structural question is whether buyers treat AI governance as part of their connectivity layer or as a separate purchase. If they keep the same control point for API, LLM, and agent traffic, Kong extends a durable position, and if they split the purchase, the AI line faces specialist rivals and platform bundling without the installed-base protection. [s1, s5, s3]

Kong's strongest go-to-market signal for the AI line is distribution rather than a named AI customer. The AI Gateway ships into an installed base of more than 700 enterprises that Kong reports for the whole company, including GSK, PayPal, and NYSE, and into a paid sales motion that already passed 100 million dollars in recurring revenue. A new product reaching that base starts from a different place than a startup hunting for its first deployment.

The whole-company figures are real and third-party reported but do not measure the AI Gateway by itself. The Stack reported the revenue, profitability, and customer count around Kong's November 2024 financing, and none of that coverage separates AI Gateway adoption from API gateway adoption. The 700 customers and the recurring revenue describe Kong, not the AI line, so they support distribution potential rather than proven AI-line traction.

The gap the public record leaves is AI-line depth: the cited record carries little customer-specific traction for the AI Gateway and no AI-line revenue figure. Until Kong reports how many of its enterprises run the AI Gateway and what the line earns, the traction case depends on the installed base and the credibility of the company carrying it, which is why the indirect signal lifts but does not max the go-to-market score. [s3, s1]

Kong's founders have already built one large company in the market adjacent to AI connectivity. Augusto Marietti, the CEO, and Marco Palladino, the CTO, founded Kong and grew it from an open-source API gateway into a business that passed 100 million dollars in recurring revenue, reached profitability, and raised at a 2 billion dollar valuation. That is a verifiable prior build at scale, not a plausible background.

The proven execution is in API connectivity rather than AI security specifically. The team's record demonstrates that it can ship infrastructure software and sell it to large enterprises, which is the harder part of the AI Gateway motion, and the AI-security capability is the newer addition layered on that base. Recording the founders' track record as a structural factor for the AI line keeps the credit honest about what it does and does not prove.

The investor base reinforces the operating credibility. Kong's Series E drew Tiger Global and Balderton alongside continued support from earlier backers, and the round was an up-round at a higher valuation, which is external validation of the company carrying the AI line even though it does not speak to the AI product directly. [s6, s3]

Kong positions the AI Gateway as a governance and control product, so assurance is the value proposition rather than an afterthought. The product controls access, data leakage, and token usage for how developers, apps, and agents consume LLMs, and the PII sanitization plugin lets platform owners enforce data protection at the platform level instead of relying on each developer to code it. That posture targets the compliance and audit needs enterprises raise when moving AI into production.

The company behind the product carries enterprise-grade procurement weight. Kong is profitable, names large regulated enterprises such as GSK and a major financial exchange among its customers, and runs a mature sales and support organization, which resolves the vendor-viability questions a one-product startup would raise. A buyer evaluating the AI Gateway inherits that posture.

What a careful buyer still needs is evidence specific to the AI line. The trust story for the company is strong, but published security attestations, independent guardrail evaluations, or named AI Gateway references would close the distance between trusting Kong and trusting the AI governance product in particular. [s1, s3]

Add analyzed competitors to compare them side by side with Kong Inc..

Kong aims the AI Gateway at enterprises moving AI from pilots into production that need one control point for model and agent traffic. The product page frames the buyer as engineering and platform teams who must govern how developers, apps, and agents consume LLMs, controlling access, data leakage, and token usage in one place. That is the same platform-owner buyer who already runs an API gateway, not a new audience Kong has to find.

The named customers describe regulated, high-stakes enterprises. The Stack reports more than 700 enterprise customers including GSK, PayPal, and NYSE, a pharmaceutical company and a financial exchange. The coverage names them at the company level rather than as AI Gateway users, so these figures describe the segment Kong can reach rather than the segment that has bought the AI line.

Kong's clearest segmentation opening is to sell the AI Gateway into its existing footprint, though the cited sources do not document an actual installed-base-first sequencing choice. For a company already operating Kong, AI governance is an extension of infrastructure the platform team maintains, which narrows the target to organizations that have already standardized on Kong and want to keep one control plane as they add model traffic.

The Kong AI Gateway is a set of governance plugins on Kong's data plane, with named controls rather than marketing generalities. The documentation describes built-in PII sanitization that detects and redacts sensitive data across 20 categories and 9 languages, running privately and self-hosted, and the plan listing names prompt guardrails and LLM access control that screen requests across model providers. A buyer inspects each control through a public documentation portal.

Kong has extended the product from model traffic to agent traffic across releases. Press coverage of the Agent Gateway launch describes it giving engineering teams a single place to govern AI traffic across LLM, Model Context Protocol, and agent-to-agent communication. That progression tracks the market shift from single model requests toward multi-step agent interactions, and it widens the surface the product governs.

The capability rests on configuration rather than a proprietary detection asset. The documentation describes governance through allow and deny lists, redaction patterns, and routing policies that the buyer configures and trusts, and the public record shows no proprietary attack corpus or model that Kong alone holds. The depth is real and inspectable, and a well-funded competitor could reproduce it by writing similar plugins, which is the structural ceiling on this advantage.

Kong's go-to-market for the AI line runs through the installed API-gateway base rather than a separate AI sales motion. The AI Gateway ships into customers that already run Kong, and the pricing page bundles AI Gateway features into the Konnect plans a buyer is already on, so reaching the AI buyer does not require a new vendor relationship. That installed base is a parent distribution channel the AI line rides, not proof the AI line itself has won buyers, and it is the reach a standalone tool cannot copy by writing software.

The public traction evidence measures the company, not the AI Gateway. The Stack reported the recurring revenue, profitability, and 700-customer count, and none of that coverage separates AI Gateway adoption from API gateway adoption. The 700 customers and the recurring revenue describe Kong as a whole, so they support distribution potential for the AI line rather than proven AI-line demand.

The gap the record leaves is a named AI Gateway customer and an AI-line revenue figure. Until Kong reports how many of its enterprises run the AI Gateway and what the line earns, the go-to-market case depends on the installed base and the credibility of the company carrying it, not on dated proof that buyers are choosing the AI Gateway specifically.

Kong publishes pricing for the platform that carries the AI Gateway, and it folds the AI line into existing plan tiers rather than charging for it separately. The pricing page lists AI Gateway features inside the Konnect Plus tier, which is charged per gateway per month, including a universal LLM API across up to 5 models, MCP server proxies, PII sanitization, prompt guardrails, and token-based rate limiting with semantic caching.

The charging unit reveals what Kong believes the buyer pays for. Konnect Plus bills by the gateway, the runtime control point, which signals that the buyer pays for a governed traffic plane rather than per seat or per AI feature. The AI-specific cost controls inside the plan, token-based rate limiting and token-level tracking, align the product with how buyers measure AI spend.

The published price covers the lower tier, and the enterprise motion stays negotiated. The Plus plan is self-serve with a public rate, while larger deployments route to sales, which fits a product sold into regulated enterprises that buy through procurement. The AI Gateway inherits this two-track model from the platform rather than carrying its own price.

The Kong AI Gateway runs wherever the Kong data plane runs, which lets the buyer keep AI traffic inside its own boundary. The documentation states that AI plugins are supported in all deployment modes, including Konnect, self-hosted traditional, hybrid, and DB-less, and on Kubernetes via the Kong Ingress Controller. A buyer can run the gateway privately and self-hosted, which the data governance documentation ties to full control and compliance.

The operational model is the same one Kong customers already run for API traffic. The AI Gateway is plugins on the existing gateway, so a team that operates Kong adds AI governance without standing up a new system, and the same declarative configuration and deployment topologies carry over. That reuse lowers the operational cost of adopting the AI line for an existing Kong operator.

The self-hosted and hybrid options matter particularly for the regulated buyers Kong targets. An enterprise that cannot send prompts and responses to a third-party SaaS can run the governance plane in its own environment, which removes a common blocker for AI adoption in pharmaceutical, financial, and government settings.

Kong positions the AI Gateway as a governance and control product, so assurance is the value proposition rather than an afterthought. The product controls access, data leakage, and token usage for how developers, apps, and agents consume LLMs, and the PII sanitization documented across 20 categories and 9 languages lets platform owners enforce data protection centrally instead of relying on each developer to code it.

The private deployment option is the strongest trust signal specific to the AI line. Because the gateway can run privately and self-hosted, an enterprise keeps prompts and responses inside its own environment, which the data governance documentation frames as full control and compliance. That answers the data-residency concern that blocks many AI deployments more directly than any attestation would.

What a careful buyer still needs is evidence specific to the AI line rather than the company. The public record carries no independent guardrail evaluation, no AI-Gateway-specific security attestation, and no named AI Gateway reference customer, so the trust case for the product itself rests on documented controls and the procurement weight of a profitable vendor rather than on third-party validation.

The AI Gateway's ecosystem advantage is that it is part of Kong's existing plugin platform, so AI routes inherit the gateway's toolset. The plan listing places the AI Gateway features inside the same plan that carries REST, gRPC, GraphQL, Kafka, and WebSocket support, so the auth, rate limiting, and routing that serve API traffic also serve LLM and agent routes on one control plane.

Kong reaches buyers through channels beyond its own sales team. The pricing page offers the Plus plan as a self-serve Start for Free option and through Buy with AWS, so a buyer can adopt the AI Gateway through infrastructure and procurement paths it already uses. That marketplace and multi-protocol reach is a distribution surface a single-purpose AI tool would have to build from scratch.

The ecosystem also covers the agent surface the market is moving toward. The product governs MCP servers and agent-to-agent traffic, extending the gateway from model calls to multi-agent workflows, which keeps the AI Gateway positioned at the connectivity layer as the unit of AI traffic shifts from a single request to a chain of agent interactions.

Kong's founders have already built one large company in the market next to AI connectivity. They grew Kong from an open-source API gateway into a business that passed 100 million dollars in recurring revenue and reached profitability, a verifiable prior build at scale rather than a plausible background. That record shows the team can ship infrastructure software and sell it to large enterprises.

The proven execution is in API connectivity rather than AI security specifically. The team's record demonstrates the harder part of the AI Gateway motion, shipping and selling enterprise infrastructure, while the AI-security capability is the newer addition layered on that base. Recording the founders' track record as a structural factor for the AI line keeps the credit honest about what it does and does not prove.

The company carries the organizational weight that the AI line inherits. Kong runs a mature sales and support organization and a documented platform, which resolves the vendor-viability questions a one-product startup would raise, so a buyer evaluating the AI Gateway inherits an established operator rather than betting on a new entrant.