Kong's SaaS API management platform designed to simplify managing and securing APIs.
Use AWS Guardrails to validate requests and/or responses before forwarding them between clients and upstream LLMs.
Use Azure AI Content Safety to check and audit messages before proxying them to an upstream LLM
Use a third-party guardrails service to validate requests and/or responses before forwarding them between clients and upstream LLMs
Audit and validate LLM prompts with Google Cloud Model Armor before forwarding them to an upstream LLM.
Score prompt-response pairs with a dedicated LLM to evaluate the accuracy and quality of model responses.
Inspect and enforce Lakera Guard safety policies on LLM requests and responses before they reach upstream models.
Check LLM requests and responses against NVIDIA NeMo Guardrails and block content that violates your safety rails.
Protect sensitive information in client request or response bodies before they reach upstream services or clients
Compress prompts before they reach the upstream LLM to stay within context limits, cut token costs, and reduce latency.
Prepend or append an array of messages to a user’s chat history.
Check text completion requests against a list of allowed or denied expressions
Provide fill-in-the-blank AI prompts to users
Create RAG pipelines by automatically injecting content from a vector database
Provides rate limiting for the AI plugins.
Use an LLM service to transform a client request body prior to proxying the request to the upstream server
Use an LLM service to transform the upstream HTTP(S) prior to forwarding it to the client
Cache LLM responses based on semantic similarity between requests to reduce latency and cost.
Permit or block LLM prompts based on semantic similarity to predefined allow and deny rules.
Permit or block LLM responses based on semantic similarity to predefined rules for chat, completions, and embeddings requests
Delegate per-request LLM model selection to the NVIDIA Switchyard Decision API, dispatched natively as an AI Gateway custom policy.
Integrate AI Gateway with the AppDynamics APM Platform
Send request metrics, such as request count, request and response size, and latency, to a Datadog agent. Covers the metrics and tags the policy sends, scopes, queuing, and limitations.
Propagate spans and report space to a backend server through OTLP protocol.
Expose metrics related to Kong AI Gateway in Prometheus exposition format
Send request metrics, such as request count, request and response size, and latency, to a StatsD server. Covers the metrics, tags, naming options, scopes, and limitations.
Propagate Zipkin spans and report tracing data to a Zipkin server
Append request and response data to a log file
Send request and response logs to an HTTP server
Publish logs to a Kafka topic.
Send request and response logs to Loggly
Publish request and response logs to a Solace endpoint or topic
Send request and response logs to Syslog
Send request and response logs to a TCP server
Send request and response logs to a UDP server
A licensed add-on through Konnect Metering & Billing. Meter API requests and AI token usage for usage-based billing. Supports flexible customer identification, custom pricing dimensions, and fine-grained traffic filtering.
Let’s Encrypt and ACMEv2 integration with Kong AI Gateway
Block crawlers, scripted clients, and other unwanted User-Agents from calling your AI Models, AI Agents, and AI MCP Servers.
Configure Cross-Origin Resource Sharing (CORS) so that browser clients on approved origins can call your AI Models, AI Agents, and AI MCP Servers.
Allow or deny client IP addresses and CIDR ranges that can send requests to your AI Models, AI Agents, and AI MCP Servers.
Detect and block requests that contain SQL, server-side include, or other injection patterns before they reach your AI Models, AI Agents, and AI MCP Servers.
Apply size checks on JSON payload and minimize risk of content-level attacks
Authorize requests against Open Policy Agent
Requests a client to present its client certificate
Proxies TLS client certificate metadata to upstream services via an HTTP headers
Invoke and manage AWS Lambda functions from Kong AI Gateway
Invoke and manage Azure functions from Kong AI Gateway
Add and manage custom Lua functions to execute after other plugins
Add and manage custom Lua functions to run before other plugins
Control which AI Consumers and AI Consumer Groups can access entities
Gradually roll out a new upstream to a subset of requests for your AI Models, AI Agents, and AI MCP Servers.
Consume messages from Confluent Cloud Kafka topics and make them available through HTTP, server-sent events, or WebSocket endpoints.
Cache and serve commonly requested responses in Kong AI Gateway
Provides rate limiting for GraphQL queries
Consume messages from Kafka topics and make them available through HTTP endpoints.
Return responses from an OpenAPI spec instead of calling an AI Model Provider, so you can exercise AI Gateway without spending tokens
Validate HTTP requests and responses based on an OpenAPI 3.0 or Swagger API Specification
Cache upstream responses in memory and serve repeated requests without calling the upstream again.
Cache upstream responses in memory or Redis and serve repeated requests from the cache for your AI Models, AI Agents, and AI MCP Servers.
Limit how many HTTP requests can be made in a given period of seconds, minutes, hours, days, months, or years.
Enhanced rate limiting capabilities, for example sliding window support, Redis Sentinel support, and increased performance.
Stop requests to your AI Models, AI Agents, and AI MCP Servers and send callers to a new URL with a redirect response.
Block requests with bodies greater than a specified size.
Terminate requests to your AI Models, AI Agents, and AI MCP Servers with a specific status code and response.
Validates requests before they reach the upstream service
Rate limit based on a custom response header value.
Send requests to a different host when the request headers match a rule, for AI Agents and AI MCP Servers.
Prevent abuse and protect services with absolute limits on the number of requests reaching the service
Consume messages from Solace topics and make them available through HTTP endpoints
Validate that incoming webhooks adhere to the Standard Webhooks specification
Set custom timeouts on connections to upstream services to override Gateway Service-level timeouts.
Block incoming WebSocket messages greater than a specified size
Validate WebSocket messages before they are proxied
Apply structural and size checks on XML payloads
Transform requests into Kafka messages in a Confluent Kafka topic.
Add a unique ID header to requests sent to your AI Models, AI Agents, and AI MCP Servers, so you can trace a request through your upstream services and logs.
Build workflows that call external APIs and use the responses to rewrite AI requests and responses, or answer the client directly
Transform a GraphQL upstream into a REST API
Customize the status code, body, and headers of responses that the data plane generates, such as authentication errors and unmatched route errors, using Lua functions.
Transform requests into Kafka messages in a Kafka topic.
Insert arbitrary API calls before proxying a request to the upstream service.
Use regular expressions, variables, and templates to transform requests
Use powerful regular expressions, variables, and templates to transform API requests
Modify the upstream response before returning it to the client
Modify the upstream response before returning it to the client, with greater customization capabilities
Change the host, port, or path of requests that AI Gateway sends to your AI Agents and AI MCP Servers, using values from the incoming request.
Transform requests into Solace messages in a Solace queue or topic
Access gRPC services through HTTP REST
Allow browser clients to call gRPC services
Transform JSON objects included in API requests or responses using jq programs