Kong
Also known as: Kong AI Gateway, Kong Agent Gateway, Kong Konnect, Kong Gateway, Kong MCP Registry
API gateway extended to agent traffic: Kong AI Gateway governs model, MCP and agent-to-agent calls from one control plane, with RAG injection, LLM-as-judge scoring, guardrails and regional hosting.
Kong built its business on API gateways and has extended that position into what it calls AI connectivity: one governance and runtime layer across APIs, events, model calls, MCP connections and agent-to-agent traffic. This record grades that agent layer, Kong AI Gateway and its MCP and A2A gateways, not the whole API platform.
Kong AI Gateway puts model providers behind a universal LLM API, with AI Proxy Advanced balancing across the models a customer configures, token-based rate limiting, semantic caching and prompt compression.
Its MCP Traffic Gateway turns any REST API into an MCP server, aggregates several servers behind one endpoint and authenticates each AI consumer with key auth or OpenID Connect, and its A2A Traffic Gateway routes agent-to-agent calls, rewriting agent cards so traffic passes through the gateway.
Enterprise plugins add RAG at the gateway, where the AI RAG Injector embeds each prompt, searches a vector database the customer loads and injects the results, and an LLM-as-judge plugin that scores responses from 1 to 100 on live traffic.
Governance runs through the same layer: prompt and semantic guards, PII sanitization and third-party guardrails from AWS, Azure, Google and Lakera, an AI audit log format, and OpenTelemetry metrics for model, MCP and A2A traffic viewable in Konnect.
Konnect's control plane can be hosted in Australia, the EU, the Middle East, the US, India or Singapore, with data planes self-hosted or on Dedicated Cloud Gateways, and Konnect supports SSO through OpenID Connect or SAML. Kong governs agent traffic; it does not run agents, keep their memory or ask a person to approve their actions. A 30-day trial is free, Plus is billed monthly with the AI Gateway at $100 a month per model proxied, and Enterprise is custom.
Vendor details
Canonical URL
https://konghq.com
Category
Agent infrastructure
Subcategory
AI and agent traffic control plane
Funding status
Private company headquartered in San Francisco, long established in API management before extending its gateway to AI and agent traffic.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
An API gateway heritage means very broad backend integration, extended to AI through LLM provider routing with load balancing and failover, native Model Context Protocol support, and an MCP Registry inside the Konnect API Service Catalog for registering, discovering and governing the MCP servers and tools agents use. Kong can generate MCP servers from existing API specifications, and Insomnia supports MCP testing.
In practice
A platform team registers every internal MCP server in one catalog so agents discover governed tools instead of engineers spinning up unmanaged servers across the business.
A security team enforces PII redaction and prompt guards at the gateway instead of trusting each model provider, keeping regulated data inside its own boundary.
A finance team caps token consumption per team at the gateway with token based rate limits and watches LLM metrics in Konnect, instead of discovering the bill at the end of the month.
Sources & related URLs
Research sources
Agentic Index coverage score
8.5 / 14 capabilities · 61%
| Integrations & Tool Calling | Full |
|---|---|
|
Any REST API can become an MCP server through the AI MCP Server entity in Kong AI Gateway's MCP Traffic Gateway, which also aggregates several MCP servers behind one endpoint for clients such as Claude Desktop and Cursor. AI consumers authenticate through an AI Auth Strategy (key auth or OpenID Connect, with OAuth2 metadata from the MCP spec), and AI Policies handle access control and observability. So agents act on the customer's systems through the gateway, with authentication. Sourcedeveloper.konghq.com/ai-gateway/mcpread 2026-09-22 |
|
| Workflow Orchestration | Not documented |
|
Agent to agent requests pass through Kong's A2A Traffic Gateway, which routes them via AI Gateway, extracts task metadata and rewrites agent cards so clients route through the gateway. The agents and their steps run on the customer's own servers. Kong has no workflow definition or agent execution of its own, nothing that sequences, branches, retries or routes work across deterministic workflow nodes and autonomous agent steps. Sourcedeveloper.konghq.com/ai-gateway/a2aread 2026-09-22 |
|
| Knowledge Grounding & RAG | Full |
|
RAG runs at the gateway through the AI RAG Injector plugin. Content is sent to a configured vector database through the Admin API or decK, and for each request the gateway embeds the prompt, queries the vector database for the most similar content and injects it before the model call, with access to the vector database limited to the gateway. New customer knowledge enters a persistent index without retraining, and that index scales past the context window and stays queryable. Some of this sits on higher tiers. Sourcedeveloper.konghq.com/plugins/ai-rag-injectorread 2026-09-22 |
|
| Human Oversight & Guardrails | Partial |
|
Safety and data loss policies apply to model traffic automatically. They include prompt guards, semantic prompt and response guards, PII sanitization, and third party guardrails from AWS, Azure, Google Model Armor and Lakera, plus token based rate limits and access control per consumer on models, MCP servers and agents. The customer controls these constraints on what an agent may do. There is no step where a person reviews, approves or must validate before an agent action commits. Sourcedeveloper.konghq.com/ai-gateway/ai-data-govread 2026-09-22 |
|
| Security, Identity & Governance | Full |
|
Access controls go deep. Konnect offers SSO through OpenID Connect or SAML with IdP attribute mapping, custom teams and roles, and Konnect and Dev Portal audit logs. On the AI Gateway itself, AI Auth Strategies, key auth or OpenID Connect, authenticate each AI consumer for models, MCP servers and agents, and Kong Gateway supports FIPS 140-3. Kong confirms no security attestation on its own pages. Sourcedeveloper.konghq.com/konnect-platform/ssoread 2026-09-22 |
|
| Observability & Auditability | Full |
|
AI traffic is recorded for the customer to review. Logs and LLM metrics can be viewed in Konnect or exported to any OpenTelemetry tool, AI policies write to a standard AI audit log format, and OpenTelemetry metrics cover generative AI, MCP and A2A traffic, with protocol aware metrics and tracing for calls between agents. Customers can inspect what agents sent and received step by step, keep an audit record separate from runtime metrics, and export it to their own tools. Sourcedeveloper.konghq.com/ai-gateway/ai-data-govread 2026-09-22 |
|
| Memory & State Persistence | Not documented |
|
The gateway governs traffic in flight. Its AI plugins and entities cover proxying, guardrails, rate limiting, caching, RAG injection, MCP and A2A routing and observability, and none persists an agent's state or a user's context across sessions. There is no memory layer with a stated scope and lifetime. Sourcekonghq.com/pricingread 2026-09-22 |
|
| Deployment & Data Residency | Full |
|
Konnect lets an organization host its control plane in a region it chooses in a region picker, Australia, EU, Middle East, US, India or Singapore, with gateway services, routes, consumers and other objects kept in that geo. Dedicated Cloud Gateways run in named AWS regions, and data planes can be self hosted in hybrid, DB-less or traditional on premises topologies. Sourcedeveloper.konghq.com/konnect-platform/geosread 2026-09-22 |
|
| Prebuilt Agents, Templates & Packs | Not documented |
|
Prebuilt assets for AI traffic come as plugins for guardrails, prompt templates and decorators, rate limiting, caching and RAG injection, which configure or constrain the customer's agents instead of doing work themselves. The MCP Registry catalogs the customer's own MCP servers. There are no prebuilt workflows, templates or role specific agents for a customer to adopt. Sourcekonghq.com/pricingread 2026-09-22 |
|
| Triggers & Channel Coverage | Not documented |
|
Every model call, MCP tool call and A2A call that passes through Kong AI Gateway is started by the caller and then routed, secured and observed in flight. No schedule, event or webhook starts an agent run, and the gateway acts only on requests a client has already sent. Sourcedeveloper.konghq.com/ai-gateway/a2aread 2026-09-22 |
|
| Model Flexibility & Routing | Full |
|
The customer chooses the model. Kong AI Gateway offers a universal LLM API across providers, with AI Proxy Advanced balancing across models the customer configures and LLM access control and authentication per consumer. The universal LLM API, token based rate limiting and semantic caching come with Plus, which bills per unique model proxied. Sourcekonghq.com/pricingread 2026-09-22 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
OpenAPI specifications cover the Gateway Admin API, the Konnect AI Gateway API and the Konnect AI Catalog API. Configuration can be declared through decK, Terraform and the Kong Operator, and a Kong Konnect MCP Server lets AI clients manage Konnect through MCP tools. Sourcedeveloper.konghq.com/llms.txtread 2026-09-22 |
|
| Testing, Debugging & Optimization | Full |
|
Responses can be scored by a dedicated judge model through Kong's AI LLM as Judge plugin, which evaluates prompt and response pairs and gives each response a score from 1 to 100, with a configurable evaluator prompt, conversation history depth and sampling rate. Paired with AI Proxy Advanced's accuracy balancer, it compares responses from at least two models on live traffic, so customers get a readable score to compare on their own traffic. Sourcedeveloper.konghq.com/plugins/ai-llm-as-judgeread 2026-09-22 |
|
| Browser & Computer Use | Not documented |
|
Browser and computer control sit outside what Kong AI Gateway does. It routes and governs model, MCP and agent traffic and runs no browser, desktop or remote computer session for an agent. No headless fetch or third party engine is wired in either. Sourcekonghq.com/products/kong-ai-gatewayread 2026-09-22 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Product updates
Kong launched Volcano, a new hosted platform for building, deploying and running AI agents and web apps, with free sign up live. It bundles durable functions that checkpoint every step and can resume for up to a year, branchable Postgres with vector search, auth, storage and messaging, plus queues and optional human approvals for coordinating several agents.
Bears on: Workflow orchestration
View sourcePricing
Plus is billed monthly per gateway, with the AI Gateway at $100 a month per model proxied. A 30 day free trial is available, and Enterprise is custom.
Plus bills each month per gateway control plane and per unique LLM model proxied, with request volume add ons. Enterprise is custom and billed annually.
What is public
Kong publishes its Plus rate card, including $100 a month per model proxied through the AI Gateway, control plane charges and volume add ons, alongside a 30 day free trial and custom Enterprise pricing.
Billing mechanics
Plus is self serve, billed monthly and also sold through AWS Marketplace. Gateway control planes carry monthly charges, the AI Gateway bills per unique LLM model proxied (prorated hourly, enforced monthly), and analytics, metering and identity bill by volume. Enterprise is annual and custom.
Cost watchouts
The AI Gateway bill grows with the number of distinct models proxied ($100 a month each on Plus), on top of gateway control plane charges and request add ons, and the Plus plan caps AI Gateway at 5 unique models. RAG injection, LLM as judge and the semantic guard plugins are AI Gateway Enterprise plugins, sold as add ons on Plus.
Variable cost rationale
On Plus the bill is metered by gateway control plane, by unique LLM model proxied through the AI Gateway at $100 a month each, and by API request volume in 1 million request increments. Analytics, metering and identity also bill by volume. Kong sits in the path of all AI traffic, so request charges track the very volume being governed, and each model added to the gateway adds a monthly charge. Plus allows at most 5 unique models, which caps the model charge on that plan, and model charges are prorated hourly. MCP server proxies and A2A agents are unlimited on Plus, so adding agents adds no charge of its own. The open source Kong Gateway is free to self host, and Enterprise moves to a custom annual contract.
Additional watchouts
The free open source gateway is not the AI Gateway, and the governance plugins that justify the purchase, such as RAG injection and semantic guards, are Enterprise plugins sold as add ons on Plus.
Overage / add-ons
AI Gateway models are prorated hourly and enforced monthly. API request volume above plan allowances is billed in 1 million request increments.
Sales call required
Mixed (some tiers require a call)
Free / trial
30 day free trial of Plus with no credit card required, and the open source Kong Gateway is free to self host
Lowest paid plan
Plus from $25 a month plus usage, billed monthly
Commercial notes
Kong is a long established API management vendor extending into AI connectivity, so it arrives with an existing enterprise commercial motion rather than a new one. It also sells the tooling for customers to meter and monetize their own AI consumption.
Key ambiguities
The full cost of an AI Gateway deployment on Plus combines control plane, request and per model charges. Enterprise terms are custom.
Cancellation / refund
Plus is billed monthly in arrears and can be canceled from the account page.
Support SLA / resale
Plus has email support with a 2 business day SLA.
Missing data
Enterprise pricing and licensing on Konnect, and the full split of AI and agent capabilities between the Konnect tiers beyond the named Enterprise AI plugins.
Related vendors
- AgentOps — Agent observability and debugging platform: open source SDKs trace…
- AgentPhone — AgentPhone gives the AI agents a team builds their own US or…
- Agno — Python agent framework and AgentOS runtime (formerly Phidata) for…
- AIsa — Resource and payment gateway for AI agents, with one key to 110+…
- Albato — Albato is a no code integration platform that sells an embedded,…
- AlphaBitCore — AI control plane for regulated financial firms: one gateway enforces…
Alternatives to Kong
The closest documented capability profiles to Kong among agent infrastructure platforms tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Arcade7.5 / 14A lighter documented profile than Kong
- LiteLLM9.5 / 14Adds documented Workflow Orchestration and Memory & State Persistence
- Portkey7.5 / 14A lighter documented profile than Kong
- Freeplay8.0 / 14Adds documented Triggers & Channel Coverage
- Hyperspell7.0 / 14Adds documented Memory & State Persistence
- Metorial8.0 / 14Adds documented Triggers & Channel Coverage
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded