diff --git a/src/core/eval.tsx b/src/core/eval.tsx
index 26cb71a1f..17f1b926b 100644
--- a/src/core/eval.tsx
+++ b/src/core/eval.tsx
@@ -744,6 +744,7 @@ export class EvalClient implements CoreEvalClient {
         evaluators: input.evaluatorIds?.map((evaluatorId) => ({ evaluatorId })),
         dataSourceConfig,
         kmsKeyArn: input.kmsKeyArn,
+        outputConfig: input.outputConfig,
       }),
     );
   }
@@ -1092,6 +1093,7 @@ export class EvalClient implements CoreEvalClient {
       dataSourceConfig,
       insights: input.insightIds.map((insightId) => ({ insightId })),
       clusteringConfig: input.clusteringConfig,
+      outputConfig: input.outputConfig,
       evaluationExecutionRoleArn: input.evaluationExecutionRoleArn,
       enableOnCreate: input.enableOnCreate ?? true,
     });
@@ -1162,6 +1164,11 @@ export class EvalClient implements CoreEvalClient {
         dataSourceConfig,
         insights,
         clusteringConfig,
+        // Not merged from `current` the way rule/insights are: GET echoes back the
+        // service-managed default group under /aws/bedrock-agentcore/evaluations/,
+        // which the write APIs reject, so re-sending it would break every update
+        // that leaves --output-config off. Matches `online-eval update`.
+        outputConfig: update.outputConfig,
         evaluationExecutionRoleArn: update.evaluationExecutionRoleArn,
       }),
     );
diff --git a/src/handlers/eval/batch-insights/batch-insights.test.tsx b/src/handlers/eval/batch-insights/batch-insights.test.tsx
index a8b6afd48..6d4d537e9 100644
--- a/src/handlers/eval/batch-insights/batch-insights.test.tsx
+++ b/src/handlers/eval/batch-insights/batch-insights.test.tsx
@@ -152,6 +152,47 @@ describe("eval batch-insights run", () => {
     });
   });
 
+  test("forwards --output-config so insight results land in the caller's log group", async () => {
+    const { core } = await run([
+      "eval",
+      "batch-insights",
+      "run",
+      "--name",
+      "insights_run",
+      "--agent",
+      "agent-1",
+      "--output-config",
+      '{"cloudWatchConfig":{"logGroupName":"/company/agent-insights","metricsNamespace":"Company/AgentInsights","resultDestination":"DEDICATED_LOG_GROUP"}}',
+      "--json",
+    ]);
+
+    expect(core.eval.calls[0]?.args[0]).toMatchObject({
+      outputConfig: {
+        cloudWatchConfig: {
+          logGroupName: "/company/agent-insights",
+          metricsNamespace: "Company/AgentInsights",
+          resultDestination: "DEDICATED_LOG_GROUP",
+        },
+      },
+    });
+  });
+
+  test("rejects malformed --output-config JSON", async () => {
+    await expect(
+      run([
+        "eval",
+        "batch-insights",
+        "run",
+        "--name",
+        "insights_run",
+        "--agent",
+        "agent-1",
+        "--output-config",
+        "{not json",
+      ]),
+    ).rejects.toThrow(/Invalid JSON for option '--output-config'/);
+  });
+
   test("rejects explicit analysis configuration with an online evaluation source", async () => {
     await expect(
       run([
diff --git a/src/handlers/eval/batch-insights/run/index.tsx b/src/handlers/eval/batch-insights/run/index.tsx
index 775e43d58..0cb34414b 100644
--- a/src/handlers/eval/batch-insights/run/index.tsx
+++ b/src/handlers/eval/batch-insights/run/index.tsx
@@ -5,6 +5,7 @@ import { createHandler, flag } from "../../../../router";
 import { JsonRendererKey } from "../../../../tui";
 import type { Core } from "../../../types";
 import { coreOptsFromCtx } from "../../../utils";
+import { BatchOutputConfig } from "../../batch-evaluation/outputConfig";
 import { SessionSource } from "../../sessionSource";
 
 const CONFIGURATION = "Configuration:";
@@ -36,6 +37,7 @@ export const createRunBatchInsightsHandler = (core: Core, io: AppIO) =>
         z.array(z.string()).optional(),
         { group: ANALYSIS },
       ),
+      ...BatchOutputConfig.flags,
     ],
     handle: async (ctx, flags) => {
       const resolver = new SourceResolver({ stdin: io.stdin });
@@ -57,6 +59,7 @@ export const createRunBatchInsightsHandler = (core: Core, io: AppIO) =>
           evaluatorIds: flags["evaluators"],
           source,
           kmsKeyArn: flags["kms-key-arn"],
+          outputConfig: await BatchOutputConfig.resolve(flags["output-config"], resolver),
         },
         coreOptsFromCtx(ctx),
       );
diff --git a/src/handlers/eval/online-insight/create/index.tsx b/src/handlers/eval/online-insight/create/index.tsx
index f3cbf4f3e..e18e48033 100644
--- a/src/handlers/eval/online-insight/create/index.tsx
+++ b/src/handlers/eval/online-insight/create/index.tsx
@@ -6,6 +6,7 @@ import { JsonRendererKey } from "../../../../tui";
 import { SourceResolver, type AppIO } from "../../../../io";
 import type { Core } from "../../../types";
 import { assertMutuallyExclusiveFlags, coreOptsFromCtx, parseJsonFlag } from "../../../utils";
+import { OnlineEvalOutputConfigFlag } from "../../online-eval/outputConfig";
 
 const BUILTIN_INSIGHT_PREFIX = "Builtin.Insight.";
 const ARN_PREFIX = "arn:";
@@ -63,6 +64,7 @@ export const createCreateOnlineInsightHandler = (core: Core, io: AppIO) =>
         "a description of the config's monitoring purpose",
         z.string().optional(),
       ),
+      ...OnlineEvalOutputConfigFlag.flags,
     ],
     handle: async (ctx, flags) => {
       for (const id of flags["insight"]) {
@@ -91,6 +93,7 @@ export const createCreateOnlineInsightHandler = (core: Core, io: AppIO) =>
         ),
         insightIds: flags["insight"],
         clusteringConfig: frequencies ? { frequencies } : undefined,
+        outputConfig: await OnlineEvalOutputConfigFlag.resolve(flags["output-config"], source),
         evaluationExecutionRoleArn: flags["role-arn"],
         enableOnCreate:
           flags["enable-on-create"] === undefined
diff --git a/src/handlers/eval/online-insight/online-insight.test.tsx b/src/handlers/eval/online-insight/online-insight.test.tsx
index 6c798ccea..180d42a74 100644
--- a/src/handlers/eval/online-insight/online-insight.test.tsx
+++ b/src/handlers/eval/online-insight/online-insight.test.tsx
@@ -380,6 +380,42 @@ describe("flag validation", () => {
     ).rejects.toThrow(/Invalid JSON for option '--data-source-config'/);
   });
 
+  test("create rejects malformed --output-config JSON", async () => {
+    await expect(
+      run([
+        "eval",
+        "online-insight",
+        "create",
+        "--name",
+        CONFIG_NAME,
+        "--agent",
+        FIXTURE_AGENT_ID,
+        "--role-arn",
+        FIXTURE_ROLE_ARN,
+        "--insight",
+        FIXTURE_INSIGHT_ID,
+        "--sampling-rate",
+        "10",
+        "--output-config",
+        "{not json",
+      ]),
+    ).rejects.toThrow(/Invalid JSON for option '--output-config'/);
+  });
+
+  test("update rejects malformed --output-config JSON", async () => {
+    await expect(
+      run([
+        "eval",
+        "online-insight",
+        "update",
+        "--id",
+        MISSING_CONFIG_ID,
+        "--output-config",
+        "{not json",
+      ]),
+    ).rejects.toThrow(/Invalid JSON for option '--output-config'/);
+  });
+
   // --json forces the headless path so the required-flag error surfaces; without
   // it a bare invocation opens the TUI under the empty-invocation middleware.
   test.each(["get", "pause", "resume", "delete"])("%s requires --id", async (command) => {
diff --git a/src/handlers/eval/online-insight/update/index.tsx b/src/handlers/eval/online-insight/update/index.tsx
index ddd55f6d7..988e59621 100644
--- a/src/handlers/eval/online-insight/update/index.tsx
+++ b/src/handlers/eval/online-insight/update/index.tsx
@@ -6,6 +6,7 @@ import { JsonRendererKey } from "../../../../tui";
 import { SourceResolver, type AppIO } from "../../../../io";
 import type { Core } from "../../../types";
 import { assertMutuallyExclusiveFlags, coreOptsFromCtx, parseJsonFlag } from "../../../utils";
+import { OnlineEvalOutputConfigFlag } from "../../online-eval/outputConfig";
 
 const BUILTIN_INSIGHT_PREFIX = "Builtin.Insight.";
 const ARN_PREFIX = "arn:";
@@ -58,6 +59,7 @@ export const createUpdateOnlineInsightHandler = (core: Core, io: AppIO) =>
         z.string().optional(),
       ),
       flag("role-arn", "replace the IAM role the online insight assumes", z.string().optional()),
+      ...OnlineEvalOutputConfigFlag.flags,
     ],
     handle: async (ctx, flags) => {
       if (flags["endpoint"] && flags["clear-endpoint"] === "true")
@@ -98,6 +100,7 @@ export const createUpdateOnlineInsightHandler = (core: Core, io: AppIO) =>
             await source.resolveText("data-source-config", flags["data-source-config"]),
           ),
           evaluationExecutionRoleArn: flags["role-arn"],
+          outputConfig: await OnlineEvalOutputConfigFlag.resolve(flags["output-config"], source),
         },
         coreOptsFromCtx(ctx),
       );
diff --git a/src/handlers/eval/types.tsx b/src/handlers/eval/types.tsx
index 0b362d526..04b3e696a 100644
--- a/src/handlers/eval/types.tsx
+++ b/src/handlers/eval/types.tsx
@@ -181,6 +181,7 @@ export type CreateOnlineInsightInput = {
   clusteringConfig?: { frequencies: ("DAILY" | "WEEKLY" | "MONTHLY")[] };
   evaluationExecutionRoleArn: string;
   enableOnCreate?: boolean;
+  outputConfig?: OnlineEvalOutputConfig;
 } & (
   | { agent: string; endpoint?: string; dataSourceConfig?: undefined }
   | { agent?: undefined; endpoint?: undefined; dataSourceConfig: DataSourceConfig }
@@ -197,6 +198,7 @@ export type UpdateOnlineInsightInput = {
   clearEndpoint?: boolean;
   dataSourceConfig?: DataSourceConfig;
   evaluationExecutionRoleArn?: string;
+  outputConfig?: OnlineEvalOutputConfig;
 };
 
 // Online insight configs are the same OnlineEvaluationConfig resource with insights
@@ -302,6 +304,7 @@ export type StartBatchInsightsInput = {
   evaluatorIds?: string[];
   source: SessionSourceValue;
   kmsKeyArn?: string;
+  outputConfig?: OutputConfig;
 };
 
 // InvokeDatasetInput is the runtime-level shape for replaying a dataset: invoke each