The full review
What is Kong?
Kong is an API gateway and connectivity platform from Kong Inc., and it now spans considerably more than a proxy. The company positions the portfolio as a unified API and AI platform, built to secure, manage, monetize and cost-control every LLM, MCP, event and API request. The lineup covers Kong Gateway, Kong Konnect, Kong AI Gateway, Kong Event Gateway, Kong Mesh, Kong Ingress Controller and the Kong Insomnia API client.
The common thread is traffic. Calls from a client to a service, from an application to a model provider, or from one agent to another pass through a Kong data plane, where authentication, rate limiting and telemetry are applied centrally rather than rebuilt inside every service.
The open source gateway at the core
Kong Gateway is the open source foundation, which the vendor calls the most widely adopted open source API gateway available. It runs on a lightweight NGINX engine that Kong reports handles more than 50,000 transactions per second per node. Configuration arrives via an admin API, a web UI, or declarative files applied through CI/CD, the APIOps pattern Kong advocates. Plugins supply authentication, security, transformation and analytics, teams can write their own against the Kong Plugin Development Kit, and Kong Ingress Controller carries the same policy model into Kubernetes.
What Kong Konnect adds
Konnect is the hosted control plane, pitched as the API and AI platform for platform builders. It layers a Service Catalog, Runtime Management, a Developer Portal, Advanced Analytics and usage billing over the same data planes, giving a platform team an inventory of every service plus the levers to govern it. Konnect also decides where the runtime lives: the plans offer Serverless Gateways, Hybrid Gateways that keep all connectivity data locked down in your own environment, and Dedicated Cloud Gateways that Kong manages in the cloud provider and region you choose. Coverage spans REST and HTTP, gRPC, GraphQL, Kafka and WebSockets.
Governing AI and agent traffic
Kong AI Gateway treats LLM calls, MCP servers and agent-to-agent messages as simply more traffic worth governing. It offers a unified LLM interface so applications can switch providers without a rewrite, plus PII sanitization, semantic prompt guards and access control. On cost, semantic caching sits alongside tracking of AI consumption, tool usage and token spend for L7 observability, and user, model and time-bound quotas enforced at the gateway build showback and chargeback for LLM, agent and MCP usage across the enterprise. Kong also generates MCP servers on top of managed APIs, publishes them through an MCP Registry, and uses Context Mesh to govern how agents reach enterprise data.
In practice a platform group registers services in the Service Catalog, applies policy, then publishes selected APIs to a Developer Portal where developers self-serve credentials and documentation. Advanced Analytics answers who called what and how often, and metering turns that usage into invoices for teams running an API as a product. Developers get Kong Insomnia for design and testing, with Git Sync and the Inso CLI folding API specs into the same review process as application code.
How the plans are structured
Konnect opens with a free trial running 30 days without a credit card, gateway limits lifted and analytics retained throughout. The Plus plan is charged per gateway on monthly billing and sets ceilings: five Serverless Gateways, two Hybrid Gateways, two Dedicated Cloud Gateways, two Developer Portals with 20 published APIs, advanced analytics on a million requests each month, RBAC, and email support carrying a two-business-day SLA. Enterprise lifts those caps, allows self-hosted gateways, and adds Audit Logs, SSO, dedicated account managers and higher SLAs on annual billing.
Who should choose Kong
Kong suits platform engineering teams standardizing dozens or hundreds of services across Kubernetes and multiple clouds, especially where an open source on-ramp matters: you can prove the gateway out before adopting Konnect without changing the data plane. It is equally strong where LLM, MCP and agent traffic should fall under the same governance as existing APIs. It is less compelling for a small team with one or two APIs and no dedicated platform capacity; the catalog, portal, runtime and plugin surface is real operational work, and a managed gateway from your cloud provider will get you there faster.
Strengths
- Free, mature open-source gateway with a huge plugin ecosystem
- Konnect adds managed control plane, analytics, and dev portal
- Native AI Gateway routes and governs multiple LLM providers
- Deploys anywhere: SaaS, hybrid, dedicated cloud, or self-hosted
Watch out for
- Enterprise pricing is quote-only with no public rate card
- Per-million-request overages ($200/1M) get costly at high volume
- Plus caps traffic at 10M requests/month, forcing Enterprise upgrades
- Steeper learning curve than fully-managed cloud-native gateways
Key features
| Feature | What it does |
|---|
| Kong Gateway | High-performance API gateway supporting REST, gRPC, GraphQL, Kafka, and WebSocket, with a broad plugin ecosystem for auth, rate limiting, and transforms. |
| AI Gateway | Routes, secures, and governs traffic across multiple LLM providers; Plus includes 5 models, with prompt management and semantic controls. |
| Kong Konnect | SaaS control plane delivering centralized management, analytics, and configuration across serverless, hybrid, and dedicated cloud gateways. |
| Developer Portal | Self-service portal for API discovery, docs, and onboarding; up to 2 portals on Plus, unlimited on Enterprise. |
| Service Catalog | Centralized inventory of services and APIs for governance and visibility across the organization. |
| Kong Mesh | Kuma-based service mesh for securing and observing service-to-service traffic in Kubernetes and multi-cloud environments. |
How Kong compares
| Alternative | How it differs |
|---|
| Google Apigee | Enterprise API management with deep analytics and monetization; pricier, more full-lifecycle, also largely quote-based. |
| AWS API Gateway | Fully managed, pay-per-request ($1-$3.50 per million); simplest for AWS-native stacks but less portable and plugin-rich. |
| Tyk | Open-source-first gateway similar to Kong; competitive self-hosted option with more transparent published pricing. |
Frequently asked questions
How much is Kong?
Kong Konnect Plus starts at $25/month for a serverless gateway, $200/month hybrid, or $500/month for a dedicated cloud gateway, each including 1 million requests. Extra traffic costs $200 per additional million (capped at 10M/month). Enterprise is custom-quoted annually, reportedly starting around $30,000-$50,000 per year.
Is Kong free?
Yes. The open-source Kong Gateway is completely free to self-host, with the core API gateway and plugin ecosystem included. Kong Konnect, the managed SaaS platform, offers a 30-day free trial of full Enterprise functionality before you move to a paid Plus or Enterprise plan.
Is Kong open source?
Yes. Kong Gateway is a widely used open-source API gateway, and the service mesh builds on the open-source Kuma project. The paid Konnect platform layers a managed control plane, analytics, developer portal, and enterprise security (SSO, RBAC, audit logs) on top of that open-source foundation.
Kong vs Apigee?
Kong is lighter, plugin-driven, and portable across cloud and self-hosted, with a free open-source core and per-gateway pricing from $25/month. Apigee is Google's heavier full-lifecycle platform with stronger built-in monetization and analytics but higher, largely quote-based pricing. Kong suits engineering-led teams; Apigee suits large API-product programs.
What is Kong Konnect?
Kong Konnect is Kong's SaaS control plane for managing API traffic across serverless, hybrid, and dedicated cloud gateways from one place. It adds centralized analytics, a developer portal, service catalog, and AI Gateway. It has Free trial, Plus, and custom Enterprise tiers, versus the free self-hosted open-source gateway.
Kong ratings on other platforms
Independent user ratings from third-party review sites, linked here for transparency.
These are not our editorial score, are captured on the date shown, and may have changed since.