koog-bedrock-agentcore-runtime

A Ktor route installer that implements the Amazon Bedrock AgentCore Runtime HTTP contract, so a Koog agent can be deployed and invoked as an AgentCore Runtime workload.

The route installer exposes the two endpoints required by the runtime:

POST /invocations — the invocation endpoint, backed by a user-supplied handler. Supports JSON-in/JSON-out (typed handle<I, O>) as well as raw text, binary, multipart, and streaming payloads (unified handler).
GET /ping — a health check that reports Healthy, HealthyBusy (while background work tracked by AgentCoreTaskTracker is in progress), or Unhealthy.

Additional capabilities:

Optional per-request handler timeout (handlerTimeoutMillis, returns 504).
Request body size limit (maxRequestBytes, returns 413).
Streaming threshold to expose large/length-less bodies as InvocationInput.Stream.
Compatibility with application- or route-level Ktor rate limiting.
Custom health checks through AgentCorePingService.

Usage

The handler runs with a Ktor RoutingContext receiver, so it composes directly with the ../koog-ktor plugin: install the Koog plugin to configure the LLM (here, Amazon Bedrock), then forward the AgentCore invocation request into aiAgent.

import ai.koog.agentcore.runtime.agentCoreRuntime
import ai.koog.agentcore.runtime.handle
import ai.koog.ktor.Koog
import ai.koog.ktor.aiAgent
import ai.koog.ktor.bedrock
import ai.koog.prompt.executor.clients.bedrock.BedrockModels
import io.ktor.server.application.Application
import io.ktor.server.application.install
import io.ktor.server.routing.routing
import kotlinx.serialization.Serializable

@Serializable
data class InvocationRequest(val prompt: String)

@Serializable
data class InvocationResponse(val answer: String)

fun Application.module() {
    // koog-ktor: configure the Amazon Bedrock LLM client (see BedrockConfig.kt)
    install(Koog) {
        bedrock(apiKey = System.getenv("AWS_BEARER_TOKEN_BEDROCK")) {
            region = "us-east-1"
        }
    }

    // koog-bedrock-agentcore-runtime: expose /invocations and /ping
    routing {
        agentCoreRuntime {
            handle<InvocationRequest, InvocationResponse> { request, ctx ->
                // `this` is RoutingContext, so aiAgent(...) from koog-ktor is available.
                // ctx exposes AgentCore headers (session id, user id, etc.).
                val answer = aiAgent(
                    input = request.prompt,
                    model = BedrockModels.AmazonNovaMicro,
                )
                InvocationResponse(answer)
            }
        }
    }
}

For non-JSON payloads, set the unified handler instead of the typed handle<I, O> and return an InvocationResult (Text, Binary, TextStream, or BinaryStream).

Packages

ai.koog.agentcore.runtime

jvm