diff --git a/src/platform/configuration/common/configurationService.ts b/src/platform/configuration/common/configurationService.ts
index bcb7c1aab..85a168fdf 100644
--- a/src/platform/configuration/common/configurationService.ts
+++ b/src/platform/configuration/common/configurationService.ts
@@ -828,6 +828,12 @@ export namespace ConfigKey {
 		export const InstantApplyModelName = defineTeamInternalSetting<string>('chat.advanced.instantApply.modelName', ConfigType.ExperimentBased, CHAT_MODEL.GPT4OPROXY);
 		export const VerifyTextDocumentChanges = defineTeamInternalSetting<boolean>('chat.advanced.inlineEdits.verifyTextDocumentChanges', ConfigType.ExperimentBased, false);
 		export const UseAutoModeRouting = defineTeamInternalSetting<boolean>('chat.advanced.useAutoModeRouter', ConfigType.ExperimentBased, false);
+		/** Controls which routing method the auto-intent-service uses per-request.
+		 * '' (empty/default) = use server default.
+		 * 'binary' = binary classifier v1 (path B).
+		 * 'hydra' = HYDRA multi-head capability matching (path C).
+		 * Used for A/B/C experiments: A=automod, B=binary, C=hydra. */
+		export const AutoModeRoutingMethod = defineTeamInternalSetting<string>('chat.advanced.autoModeRoutingMethod', ConfigType.ExperimentBased, '', undefined, undefined, { experimentName: 'copilotchat.autoModeRoutingMethod' });
 
 		/** Inline Completions */
 		export const InlineCompletionsDefaultDiagnosticsOptions = defineTeamInternalSetting<string | undefined>('chat.advanced.inlineCompletions.defaultDiagnosticsOptionsString', ConfigType.ExperimentBased, undefined);
diff --git a/src/platform/endpoint/node/automodeService.ts b/src/platform/endpoint/node/automodeService.ts
index 830f91f3c..fb3810ff5 100644
--- a/src/platform/endpoint/node/automodeService.ts
+++ b/src/platform/endpoint/node/automodeService.ts
@@ -324,7 +324,13 @@ export class AutomodeService extends Disposable implements IAutomodeService {
 				previous_model: entry?.endpoint?.model,
 				turn_number: (entry?.turnCount ?? 0) + 1,
 			};
-			const result = await this._routerDecisionFetcher.getRouterDecision(prompt, token.session_token, token.available_models, undefined, contextSignals, chatRequest?.sessionId, chatRequest?.id);
+			const routingMethod = this._configurationService.getExperimentBasedConfig(ConfigKey.TeamInternal.AutoModeRoutingMethod, this._expService) || undefined;
+			const result = await this._routerDecisionFetcher.getRouterDecision(prompt, token.session_token, token.available_models, undefined, contextSignals, chatRequest?.sessionId, chatRequest?.id, routingMethod);
+
+			if (result.fallback) {
+				this._logService.info(`[AutomodeService] Router signaled fallback: ${result.fallback_reason ?? 'unknown'}, routing_method=${result.routing_method ?? 'n/a'}`);
+				return { lastRoutedPrompt: prompt, fallbackReason: `routerFallback:${result.fallback_reason ?? 'unknown'}` };
+			}
 
 			if (!result.candidate_models.length) {
 				return { lastRoutedPrompt: prompt, fallbackReason: 'emptyCandidateList' };
diff --git a/src/platform/endpoint/node/routerDecisionFetcher.ts b/src/platform/endpoint/node/routerDecisionFetcher.ts
index 948d68366..b98754ee6 100644
--- a/src/platform/endpoint/node/routerDecisionFetcher.ts
+++ b/src/platform/endpoint/node/routerDecisionFetcher.ts
@@ -13,7 +13,7 @@ import { ITelemetryService } from '../../telemetry/common/telemetry';
 import { ICAPIClientService } from '../common/capiClient';
 
 export interface RouterDecisionResponse {
-	predicted_label: 'needs_reasoning' | 'no_reasoning';
+	predicted_label: 'needs_reasoning' | 'no_reasoning' | 'fallback';
 	confidence: number;
 	latency_ms: number;
 	candidate_models: string[];
@@ -22,6 +22,12 @@ export interface RouterDecisionResponse {
 		no_reasoning: number;
 	};
 	sticky_override?: boolean;
+	routing_method?: string;
+	fallback?: boolean;
+	fallback_reason?: string;
+	hydra_scores?: Record<string, number>;
+	chosen_model?: string;
+	chosen_shortfall?: number;
 }
 
 export interface RoutingContextSignals {
@@ -48,7 +54,7 @@ export class RouterDecisionFetcher {
 	) {
 	}
 
-	async getRouterDecision(query: string, autoModeToken: string, availableModels: string[], stickyThreshold?: number, contextSignals?: RoutingContextSignals, conversationId?: string, vscodeRequestId?: string): Promise<RouterDecisionResponse> {
+	async getRouterDecision(query: string, autoModeToken: string, availableModels: string[], stickyThreshold?: number, contextSignals?: RoutingContextSignals, conversationId?: string, vscodeRequestId?: string, routingMethod?: string): Promise<RouterDecisionResponse> {
 		const startTime = Date.now();
 		const requestBody: Record<string, unknown> = { prompt: query, available_models: availableModels, ...contextSignals };
 		if (stickyThreshold !== undefined) {
@@ -79,7 +85,7 @@ export class RouterDecisionFetcher {
 		const text = await response.text();
 		const result: RouterDecisionResponse = JSON.parse(text);
 		const e2eLatencyMs = Date.now() - startTime;
-		this._logService.trace(`[RouterDecisionFetcher] Prediction: ${result.predicted_label}, (confidence: ${(result.confidence * 100).toFixed(1)}%, scores: needs_reasoning=${(result.scores.needs_reasoning * 100).toFixed(1)}%, no_reasoning=${(result.scores.no_reasoning * 100).toFixed(1)}%) (latency_ms: ${result.latency_ms}, e2e_latency_ms: ${e2eLatencyMs}, candidate models: ${result.candidate_models.join(', ')}, sticky_override: ${result.sticky_override ?? false})`);
+		this._logService.trace(`[RouterDecisionFetcher] Prediction: ${result.predicted_label}, (confidence: ${(result.confidence * 100).toFixed(1)}%, scores: needs_reasoning=${(result.scores.needs_reasoning * 100).toFixed(1)}%, no_reasoning=${(result.scores.no_reasoning * 100).toFixed(1)}%) (latency_ms: ${result.latency_ms}, e2e_latency_ms: ${e2eLatencyMs}, candidate models: ${result.candidate_models.join(', ')}, sticky_override: ${result.sticky_override ?? false}, routing_method: ${result.routing_method ?? 'n/a'}, fallback: ${result.fallback ?? false})`);
 
 		this._requestLogger.addEntry({
 			type: LoggedRequestKind.MarkdownContentRequest,