diff --git a/extensions/amazon-bedrock/register.runtime.ts b/extensions/amazon-bedrock/register.runtime.ts
index fbc552f9ae2..d6e72731249 100644
--- a/extensions/amazon-bedrock/register.runtime.ts
+++ b/extensions/amazon-bedrock/register.runtime.ts
@@ -46,6 +46,11 @@ function createGuardrailWrapStreamFn(
 
 const PROVIDER_ID = "amazon-bedrock";
 const CLAUDE_46_MODEL_RE = /claude-(?:opus|sonnet)-4(?:\.|-)6(?:$|[-.])/i;
+const BEDROCK_CONTEXT_OVERFLOW_PATTERNS = [
+  /ValidationException.*(?:input is too long|max input token|input token.*exceed)/i,
+  /ValidationException.*(?:exceeds? the (?:maximum|max) (?:number of )?(?:input )?tokens)/i,
+  /ModelStreamErrorException.*(?:Input is too long|too many input tokens)/i,
+] as const;
 
 export async function registerAmazonBedrockPlugin(api: OpenClawPluginApi): Promise<void> {
   const guardrail = (api.pluginConfig as Record<string, unknown> | undefined)?.guardrail as
@@ -86,6 +91,17 @@ export async function registerAmazonBedrockPlugin(api: OpenClawPluginApi): Promi
     resolveConfigApiKey: ({ env }) => resolveBedrockConfigApiKey(env),
     buildReplayPolicy: ({ modelId }) => buildAnthropicReplayPolicyForModel(modelId),
     wrapStreamFn,
+    matchesContextOverflowError: ({ errorMessage }) =>
+      BEDROCK_CONTEXT_OVERFLOW_PATTERNS.some((pattern) => pattern.test(errorMessage)),
+    classifyFailoverReason: ({ errorMessage }) => {
+      if (/ThrottlingException|Too many concurrent requests/i.test(errorMessage)) {
+        return "rate_limit";
+      }
+      if (/ModelNotReadyException/i.test(errorMessage)) {
+        return "overloaded";
+      }
+      return undefined;
+    },
     resolveDefaultThinkingLevel: ({ modelId }) =>
       CLAUDE_46_MODEL_RE.test(modelId.trim()) ? "adaptive" : undefined,
   });
diff --git a/extensions/anthropic/config-defaults.ts b/extensions/anthropic/config-defaults.ts
new file mode 100644
index 00000000000..b84c04f3eb1
--- /dev/null
+++ b/extensions/anthropic/config-defaults.ts
@@ -0,0 +1,225 @@
+import type { OpenClawConfig } from "openclaw/plugin-sdk/plugin-entry";
+import { normalizeProviderId } from "openclaw/plugin-sdk/provider-model-shared";
+
+const ANTHROPIC_PROVIDER_API = "anthropic-messages";
+
+function resolveAnthropicDefaultAuthMode(
+  config: OpenClawConfig,
+  env: NodeJS.ProcessEnv,
+): "api_key" | "oauth" | null {
+  const profiles = config.auth?.profiles ?? {};
+  const anthropicProfiles = Object.entries(profiles).filter(
+    ([, profile]) => profile?.provider === "anthropic",
+  );
+
+  const order = config.auth?.order?.anthropic ?? [];
+  for (const profileId of order) {
+    const entry = profiles[profileId];
+    if (!entry || entry.provider !== "anthropic") {
+      continue;
+    }
+    if (entry.mode === "api_key") {
+      return "api_key";
+    }
+    if (entry.mode === "oauth" || entry.mode === "token") {
+      return "oauth";
+    }
+  }
+
+  const hasApiKey = anthropicProfiles.some(([, profile]) => profile?.mode === "api_key");
+  const hasOauth = anthropicProfiles.some(
+    ([, profile]) => profile?.mode === "oauth" || profile?.mode === "token",
+  );
+  if (hasApiKey && !hasOauth) {
+    return "api_key";
+  }
+  if (hasOauth && !hasApiKey) {
+    return "oauth";
+  }
+
+  if (env.ANTHROPIC_OAUTH_TOKEN?.trim()) {
+    return "oauth";
+  }
+  if (env.ANTHROPIC_API_KEY?.trim()) {
+    return "api_key";
+  }
+  return null;
+}
+
+function resolveModelPrimaryValue(
+  value: string | { primary?: string; fallbacks?: string[] } | undefined,
+): string | undefined {
+  if (typeof value === "string") {
+    const trimmed = value.trim();
+    return trimmed || undefined;
+  }
+  const primary = value?.primary;
+  if (typeof primary !== "string") {
+    return undefined;
+  }
+  const trimmed = primary.trim();
+  return trimmed || undefined;
+}
+
+function resolveAnthropicPrimaryModelRef(raw?: string): string | null {
+  if (!raw) {
+    return null;
+  }
+  const trimmed = raw.trim();
+  if (!trimmed) {
+    return null;
+  }
+  const aliasKey = trimmed.toLowerCase();
+  if (aliasKey === "opus") {
+    return "anthropic/claude-opus-4-6";
+  }
+  if (aliasKey === "sonnet") {
+    return "anthropic/claude-sonnet-4-6";
+  }
+  return trimmed;
+}
+
+function parseProviderModelRef(
+  raw: string,
+  defaultProvider: string,
+): { provider: string; model: string } | null {
+  const trimmed = raw.trim();
+  if (!trimmed) {
+    return null;
+  }
+  const slashIndex = trimmed.indexOf("/");
+  if (slashIndex <= 0) {
+    return { provider: defaultProvider, model: trimmed };
+  }
+  const provider = trimmed.slice(0, slashIndex).trim();
+  const model = trimmed.slice(slashIndex + 1).trim();
+  if (!provider || !model) {
+    return null;
+  }
+  return {
+    provider: normalizeProviderId(provider),
+    model,
+  };
+}
+
+function isAnthropicCacheRetentionTarget(
+  parsed: { provider: string; model: string } | null | undefined,
+) {
+  return Boolean(
+    parsed &&
+    (parsed.provider === "anthropic" ||
+      (parsed.provider === "amazon-bedrock" &&
+        parsed.model.toLowerCase().includes("anthropic.claude"))),
+  );
+}
+
+export function normalizeAnthropicProviderConfig<T extends { api?: string; models?: unknown[] }>(
+  providerConfig: T,
+): T {
+  if (
+    providerConfig.api ||
+    !Array.isArray(providerConfig.models) ||
+    providerConfig.models.length === 0
+  ) {
+    return providerConfig;
+  }
+  return { ...providerConfig, api: ANTHROPIC_PROVIDER_API };
+}
+
+export function applyAnthropicConfigDefaults(params: {
+  config: OpenClawConfig;
+  env: NodeJS.ProcessEnv;
+}): OpenClawConfig {
+  const defaults = params.config.agents?.defaults;
+  if (!defaults) {
+    return params.config;
+  }
+
+  const authMode = resolveAnthropicDefaultAuthMode(params.config, params.env);
+  if (!authMode) {
+    return params.config;
+  }
+
+  let mutated = false;
+  const nextDefaults = { ...defaults };
+  const contextPruning = defaults.contextPruning ?? {};
+  const heartbeat = defaults.heartbeat ?? {};
+
+  if (defaults.contextPruning?.mode === undefined) {
+    nextDefaults.contextPruning = {
+      ...contextPruning,
+      mode: "cache-ttl",
+      ttl: defaults.contextPruning?.ttl ?? "1h",
+    };
+    mutated = true;
+  }
+
+  if (defaults.heartbeat?.every === undefined) {
+    nextDefaults.heartbeat = {
+      ...heartbeat,
+      every: authMode === "oauth" ? "1h" : "30m",
+    };
+    mutated = true;
+  }
+
+  if (authMode === "api_key") {
+    const nextModels = defaults.models ? { ...defaults.models } : {};
+    let modelsMutated = false;
+
+    for (const [key, entry] of Object.entries(nextModels)) {
+      const parsed = parseProviderModelRef(key, "anthropic");
+      if (!isAnthropicCacheRetentionTarget(parsed)) {
+        continue;
+      }
+      const current = entry ?? {};
+      const paramsValue = (current as { params?: Record<string, unknown> }).params ?? {};
+      if (typeof paramsValue.cacheRetention === "string") {
+        continue;
+      }
+      nextModels[key] = {
+        ...(current as Record<string, unknown>),
+        params: { ...paramsValue, cacheRetention: "short" },
+      };
+      modelsMutated = true;
+    }
+
+    const primary = resolveAnthropicPrimaryModelRef(
+      resolveModelPrimaryValue(
+        defaults.model as string | { primary?: string; fallbacks?: string[] } | undefined,
+      ),
+    );
+    if (primary) {
+      const parsedPrimary = parseProviderModelRef(primary, "anthropic");
+      if (isAnthropicCacheRetentionTarget(parsedPrimary)) {
+        const key = `${parsedPrimary.provider}/${parsedPrimary.model}`;
+        const entry = nextModels[key];
+        const current = entry ?? {};
+        const paramsValue = (current as { params?: Record<string, unknown> }).params ?? {};
+        if (typeof paramsValue.cacheRetention !== "string") {
+          nextModels[key] = {
+            ...(current as Record<string, unknown>),
+            params: { ...paramsValue, cacheRetention: "short" },
+          };
+          modelsMutated = true;
+        }
+      }
+    }
+
+    if (modelsMutated) {
+      nextDefaults.models = nextModels;
+      mutated = true;
+    }
+  }
+
+  if (!mutated) {
+    return params.config;
+  }
+
+  return {
+    ...params.config,
+    agents: {
+      ...params.config.agents,
+      defaults: nextDefaults,
+    },
+  };
+}
diff --git a/extensions/anthropic/index.test.ts b/extensions/anthropic/index.test.ts
index 94e777fbe46..339aef8ec35 100644
--- a/extensions/anthropic/index.test.ts
+++ b/extensions/anthropic/index.test.ts
@@ -35,4 +35,51 @@ describe("anthropic provider replay hooks", () => {
       dropThinkingBlocks: true,
     });
   });
+
+  it("defaults provider api through plugin config normalization", () => {
+    const provider = registerSingleProviderPlugin(anthropicPlugin);
+
+    expect(
+      provider.normalizeConfig?.({
+        provider: "anthropic",
+        providerConfig: {
+          models: [{ id: "claude-sonnet-4-6", name: "Claude Sonnet 4.6" }],
+        },
+      } as never),
+    ).toMatchObject({
+      api: "anthropic-messages",
+    });
+  });
+
+  it("applies Anthropic pruning defaults through plugin hooks", () => {
+    const provider = registerSingleProviderPlugin(anthropicPlugin);
+
+    const next = provider.applyConfigDefaults?.({
+      provider: "anthropic",
+      env: {},
+      config: {
+        auth: {
+          profiles: {
+            "anthropic:api": { provider: "anthropic", mode: "api_key" },
+          },
+        },
+        agents: {
+          defaults: {
+            model: { primary: "anthropic/claude-opus-4-5" },
+          },
+        },
+      },
+    } as never);
+
+    expect(next?.agents?.defaults?.contextPruning).toMatchObject({
+      mode: "cache-ttl",
+      ttl: "1h",
+    });
+    expect(next?.agents?.defaults?.heartbeat).toMatchObject({
+      every: "30m",
+    });
+    expect(
+      next?.agents?.defaults?.models?.["anthropic/claude-opus-4-5"]?.params?.cacheRetention,
+    ).toBe("short");
+  });
 });
diff --git a/extensions/anthropic/register.runtime.ts b/extensions/anthropic/register.runtime.ts
index 964ba629289..c9f0f380d0a 100644
--- a/extensions/anthropic/register.runtime.ts
+++ b/extensions/anthropic/register.runtime.ts
@@ -30,6 +30,10 @@ import { composeProviderStreamWrappers } from "openclaw/plugin-sdk/provider-stre
 import { fetchClaudeUsage } from "openclaw/plugin-sdk/provider-usage";
 import { buildAnthropicCliBackend } from "./cli-backend.js";
 import { buildAnthropicCliMigrationResult, hasClaudeCliAuth } from "./cli-migration.js";
+import {
+  applyAnthropicConfigDefaults,
+  normalizeAnthropicProviderConfig,
+} from "./config-defaults.js";
 import { anthropicMediaUnderstandingProvider } from "./media-understanding-provider.js";
 import { buildAnthropicReplayPolicy } from "./replay-policy.js";
 import {
@@ -445,6 +449,8 @@ export async function registerAnthropicPlugin(api: OpenClawPluginApi): Promise<v
         },
       }),
     ],
+    normalizeConfig: ({ providerConfig }) => normalizeAnthropicProviderConfig(providerConfig),
+    applyConfigDefaults: ({ config, env }) => applyAnthropicConfigDefaults({ config, env }),
     resolveDynamicModel: (ctx) => resolveAnthropicForwardCompatModel(ctx),
     buildReplayPolicy: (ctx) => buildAnthropicReplayPolicy(ctx),
     isModernModelRef: ({ modelId }) => matchesAnthropicModernModel(modelId),
diff --git a/extensions/cloudflare-ai-gateway/index.ts b/extensions/cloudflare-ai-gateway/index.ts
index 11a2dce9274..15f758c0740 100644
--- a/extensions/cloudflare-ai-gateway/index.ts
+++ b/extensions/cloudflare-ai-gateway/index.ts
@@ -246,6 +246,8 @@ export default definePluginEntry({
           return null;
         },
       },
+      classifyFailoverReason: ({ errorMessage }) =>
+        /\bworkers?_ai\b.*\b(?:rate|limit|quota)\b/i.test(errorMessage) ? "rate_limit" : undefined,
     });
   },
 });
diff --git a/extensions/deepseek/index.ts b/extensions/deepseek/index.ts
index cc8086e23c4..f071942df99 100644
--- a/extensions/deepseek/index.ts
+++ b/extensions/deepseek/index.ts
@@ -34,5 +34,7 @@ export default defineSingleProviderPluginEntry({
     catalog: {
       buildProvider: buildDeepSeekProvider,
     },
+    matchesContextOverflowError: ({ errorMessage }) =>
+      /\bdeepseek\b.*(?:input.*too long|context.*exceed)/i.test(errorMessage),
   },
 });
diff --git a/extensions/mistral/index.ts b/extensions/mistral/index.ts
index acec875b306..402e1592e4a 100644
--- a/extensions/mistral/index.ts
+++ b/extensions/mistral/index.ts
@@ -89,6 +89,8 @@ export default defineSingleProviderPluginEntry({
       buildProvider: buildMistralProvider,
       allowExplicitBaseUrl: true,
     },
+    matchesContextOverflowError: ({ errorMessage }) =>
+      /\bmistral\b.*(?:input.*too long|token limit.*exceeded)/i.test(errorMessage),
     normalizeResolvedModel: ({ model }) => applyMistralModelCompat(model),
     contributeResolvedModelCompat: ({ modelId, model }) =>
       shouldContributeMistralCompat({ modelId, model }) ? MISTRAL_MODEL_COMPAT_PATCH : undefined,
diff --git a/extensions/ollama/index.ts b/extensions/ollama/index.ts
index a7e40f47ef7..7c4de35cbb6 100644
--- a/extensions/ollama/index.ts
+++ b/extensions/ollama/index.ts
@@ -168,6 +168,9 @@ export default definePluginEntry({
           client,
         };
       },
+      matchesContextOverflowError: ({ errorMessage }) =>
+        /\bollama\b.*(?:context length|too many tokens|context window)/i.test(errorMessage) ||
+        /\btruncating input\b.*\btoo long\b/i.test(errorMessage),
       resolveSyntheticAuth: ({ providerConfig }) => {
         const hasApiConfig =
           Boolean(providerConfig?.api?.trim()) ||
diff --git a/extensions/openai/openai-provider.ts b/extensions/openai/openai-provider.ts
index ebce103429c..7f4af5f87e4 100644
--- a/extensions/openai/openai-provider.ts
+++ b/extensions/openai/openai-provider.ts
@@ -259,6 +259,8 @@ export function buildOpenAIProvider(): ProviderPlugin {
       normalizeProviderId(ctx.provider) === PROVIDER_ID
         ? wrapOpenAIProviderStream(ctx)
         : wrapAzureOpenAIProviderStream(ctx),
+    matchesContextOverflowError: ({ errorMessage }) =>
+      /content_filter.*(?:prompt|input).*(?:too long|exceed)/i.test(errorMessage),
     resolveTransportTurnState: (ctx) => resolveOpenAITransportTurnState(ctx),
     resolveWebSocketSessionPolicy: (ctx) => resolveOpenAIWebSocketSessionPolicy(ctx),
     resolveReasoningOutputMode: () => "native",
diff --git a/extensions/together/index.ts b/extensions/together/index.ts
index 42ecad1a3a2..92dca5f5dda 100644
--- a/extensions/together/index.ts
+++ b/extensions/together/index.ts
@@ -30,5 +30,9 @@ export default defineSingleProviderPluginEntry({
     catalog: {
       buildProvider: buildTogetherProvider,
     },
+    classifyFailoverReason: ({ errorMessage }) =>
+      /\bconcurrency limit\b.*\b(?:breached|reached)\b/i.test(errorMessage)
+        ? "rate_limit"
+        : undefined,
   },
 });
diff --git a/src/agents/pi-embedded-helpers/provider-error-patterns.test.ts b/src/agents/pi-embedded-helpers/provider-error-patterns.test.ts
index ca4db4dfd19..5848ed244b1 100644
--- a/src/agents/pi-embedded-helpers/provider-error-patterns.test.ts
+++ b/src/agents/pi-embedded-helpers/provider-error-patterns.test.ts
@@ -59,6 +59,12 @@ describe("classifyProviderSpecificError", () => {
     expect(classifyProviderSpecificError("concurrency limit reached")).toBe("rate_limit");
   });
 
+  it("classifies Cloudflare Workers AI quota errors as rate_limit", () => {
+    expect(classifyProviderSpecificError("workers_ai gateway error: quota limit exceeded")).toBe(
+      "rate_limit",
+    );
+  });
+
   it("does not match generic 'model is not ready' without Bedrock prefix", () => {
     expect(classifyProviderSpecificError("model is not ready")).toBeNull();
   });
diff --git a/src/agents/pi-embedded-helpers/provider-error-patterns.ts b/src/agents/pi-embedded-helpers/provider-error-patterns.ts
index 391d727c717..233a34a5bbe 100644
--- a/src/agents/pi-embedded-helpers/provider-error-patterns.ts
+++ b/src/agents/pi-embedded-helpers/provider-error-patterns.ts
@@ -1,11 +1,15 @@
 /**
- * Provider-specific error patterns that improve failover classification accuracy.
+ * Provider-owned error-pattern dispatch plus legacy fallback patterns.
  *
- * Many providers return errors in non-standard formats. Without these patterns,
- * errors get misclassified (e.g., a context overflow classified as "format"),
- * causing the failover engine to choose wrong recovery strategies.
+ * Most provider-specific failover classification now lives on provider-plugin
+ * hooks. This module keeps only fallback patterns for providers that do not
+ * yet ship a dedicated provider plugin hook surface.
  */
 
+import {
+  classifyProviderFailoverReasonWithPlugin,
+  matchesProviderContextOverflowWithPlugin,
+} from "../../plugins/provider-runtime.js";
 import type { FailoverReason } from "./types.js";
 
 type ProviderErrorPattern = {
@@ -21,30 +25,9 @@ type ProviderErrorPattern = {
  * to catch provider-specific wording that the generic regex misses.
  */
 export const PROVIDER_CONTEXT_OVERFLOW_PATTERNS: readonly RegExp[] = [
-  // AWS Bedrock
-  /ValidationException.*(?:input is too long|max input token|input token.*exceed)/i,
-  /ValidationException.*(?:exceeds? the (?:maximum|max) (?:number of )?(?:input )?tokens)/i,
-  /ModelStreamErrorException.*(?:Input is too long|too many input tokens)/i,
-
-  // Azure OpenAI (sometimes wraps OpenAI errors differently)
-  /content_filter.*(?:prompt|input).*(?:too long|exceed)/i,
-
-  // Ollama / local models
-  /\bollama\b.*(?:context length|too many tokens|context window)/i,
-  /\btruncating input\b.*\btoo long\b/i,
-
-  // Mistral
-  /\bmistral\b.*(?:input.*too long|token limit.*exceeded)/i,
-
-  // Cohere
+  // Cohere does not currently ship a bundled provider hook.
   /\btotal tokens?.*exceeds? (?:the )?(?:model(?:'s)? )?(?:max|maximum|limit)/i,
 
-  // DeepSeek
-  /\bdeepseek\b.*(?:input.*too long|context.*exceed)/i,
-
-  // Google Vertex / Gemini: INVALID_ARGUMENT with token-related messages is context overflow.
-  /INVALID_ARGUMENT.*(?:exceeds? the (?:maximum|max)|input.*too (?:long|large))/i,
-
   // Generic "input too long" pattern that isn't covered by existing checks
   /\binput (?:is )?too long for (?:the )?model\b/i,
 ];
@@ -55,38 +38,11 @@ export const PROVIDER_CONTEXT_OVERFLOW_PATTERNS: readonly RegExp[] = [
  * produce wrong results for specific providers.
  */
 export const PROVIDER_SPECIFIC_PATTERNS: readonly ProviderErrorPattern[] = [
-  // AWS Bedrock: ThrottlingException is rate limit
-  {
-    test: /ThrottlingException|Too many concurrent requests/i,
-    reason: "rate_limit",
-  },
-
-  // AWS Bedrock: ModelNotReadyException (require class prefix to avoid false positives)
-  {
-    test: /ModelNotReadyException/i,
-    reason: "overloaded",
-  },
-
-  // Azure: content_policy_violation should not trigger failover
-  // (it's a content moderation rejection, not a transient error)
-
-  // Groq: model_deactivated is permanent
+  // Groq does not currently ship a bundled provider hook.
   {
     test: /model(?:_is)?_deactivated|model has been deactivated/i,
     reason: "model_not_found",
   },
-
-  // Together AI / Fireworks: specific rate limit messages
-  {
-    test: /\bconcurrency limit\b.*\breached\b/i,
-    reason: "rate_limit",
-  },
-
-  // Cloudflare Workers AI
-  {
-    test: /\bworkers?_ai\b.*\b(?:rate|limit|quota)\b/i,
-    reason: "rate_limit",
-  },
 ];
 
 /**
@@ -94,7 +50,11 @@ export const PROVIDER_SPECIFIC_PATTERNS: readonly ProviderErrorPattern[] = [
  * Called from `isContextOverflowError()` to catch provider-specific wording.
  */
 export function matchesProviderContextOverflow(errorMessage: string): boolean {
-  return PROVIDER_CONTEXT_OVERFLOW_PATTERNS.some((pattern) => pattern.test(errorMessage));
+  return (
+    matchesProviderContextOverflowWithPlugin({
+      context: { errorMessage },
+    }) || PROVIDER_CONTEXT_OVERFLOW_PATTERNS.some((pattern) => pattern.test(errorMessage))
+  );
 }
 
 /**
@@ -102,6 +62,12 @@ export function matchesProviderContextOverflow(errorMessage: string): boolean {
  * Returns null if no provider-specific pattern matches (fall through to generic classification).
  */
 export function classifyProviderSpecificError(errorMessage: string): FailoverReason | null {
+  const pluginReason = classifyProviderFailoverReasonWithPlugin({
+    context: { errorMessage },
+  });
+  if (pluginReason) {
+    return pluginReason;
+  }
   for (const pattern of PROVIDER_SPECIFIC_PATTERNS) {
     if (pattern.test.test(errorMessage)) {
       return pattern.reason;
diff --git a/src/config/defaults.ts b/src/config/defaults.ts
index 21ad26a2156..2b36d325353 100644
--- a/src/config/defaults.ts
+++ b/src/config/defaults.ts
@@ -1,7 +1,8 @@
 import { DEFAULT_CONTEXT_TOKENS } from "../agents/defaults.js";
-import { normalizeProviderId, parseModelRef } from "../agents/model-selection.js";
+import { normalizeProviderId } from "../agents/model-selection.js";
+import { normalizeProviderSpecificConfig } from "../agents/models-config.providers.policy.js";
+import { applyProviderConfigDefaultsWithPlugin } from "../plugins/provider-runtime.js";
 import { DEFAULT_AGENT_MAX_CONCURRENT, DEFAULT_SUBAGENT_MAX_CONCURRENT } from "./agent-limits.js";
-import { resolveAgentModelPrimaryValue } from "./model-input.js";
 import {
   LEGACY_TALK_PROVIDER_ID,
   normalizeTalkConfig,
@@ -16,8 +17,6 @@ type WarnState = { warned: boolean };
 
 let defaultWarnState: WarnState = { warned: false };
 
-type AnthropicAuthDefaultsMode = "api_key" | "oauth";
-
 const DEFAULT_MODEL_ALIASES: Readonly<Record<string, string>> = {
   // Anthropic (pi-ai catalog uses "latest" ids without date suffix)
   opus: "anthropic/claude-opus-4-6",
@@ -54,16 +53,6 @@ const MISTRAL_SAFE_MAX_TOKENS_BY_MODEL = {
 type ModelDefinitionLike = Partial<ModelDefinitionConfig> &
   Pick<ModelDefinitionConfig, "id" | "name">;
 
-function resolveDefaultProviderApi(
-  providerId: string,
-  providerApi: ModelDefinitionConfig["api"] | undefined,
-): ModelDefinitionConfig["api"] | undefined {
-  if (providerApi) {
-    return providerApi;
-  }
-  return normalizeProviderId(providerId) === "anthropic" ? "anthropic-messages" : undefined;
-}
-
 function isPositiveNumber(value: unknown): value is number {
   return typeof value === "number" && Number.isFinite(value) && value > 0;
 }
@@ -98,58 +87,6 @@ export function resolveNormalizedProviderModelMaxTokens(params: {
   return Math.min(safeMaxTokens, params.contextWindow);
 }
 
-function resolveAnthropicDefaultAuthMode(cfg: OpenClawConfig): AnthropicAuthDefaultsMode | null {
-  const profiles = cfg.auth?.profiles ?? {};
-  const anthropicProfiles = Object.entries(profiles).filter(
-    ([, profile]) => profile?.provider === "anthropic",
-  );
-
-  const order = cfg.auth?.order?.anthropic ?? [];
-  for (const profileId of order) {
-    const entry = profiles[profileId];
-    if (!entry || entry.provider !== "anthropic") {
-      continue;
-    }
-    if (entry.mode === "api_key") {
-      return "api_key";
-    }
-    if (entry.mode === "oauth" || entry.mode === "token") {
-      return "oauth";
-    }
-  }
-
-  const hasApiKey = anthropicProfiles.some(([, profile]) => profile?.mode === "api_key");
-  const hasOauth = anthropicProfiles.some(
-    ([, profile]) => profile?.mode === "oauth" || profile?.mode === "token",
-  );
-  if (hasApiKey && !hasOauth) {
-    return "api_key";
-  }
-  if (hasOauth && !hasApiKey) {
-    return "oauth";
-  }
-
-  if (process.env.ANTHROPIC_OAUTH_TOKEN?.trim()) {
-    return "oauth";
-  }
-  if (process.env.ANTHROPIC_API_KEY?.trim()) {
-    return "api_key";
-  }
-  return null;
-}
-
-function resolvePrimaryModelRef(raw?: string): string | null {
-  if (!raw || typeof raw !== "string") {
-    return null;
-  }
-  const trimmed = raw.trim();
-  if (!trimmed) {
-    return null;
-  }
-  const aliasKey = trimmed.toLowerCase();
-  return DEFAULT_MODEL_ALIASES[aliasKey] ?? trimmed;
-}
-
 export type SessionDefaultsOptions = {
   warn?: (message: string) => void;
   warnState?: WarnState;
@@ -242,15 +179,19 @@ export function applyModelDefaults(cfg: OpenClawConfig): OpenClawConfig {
   if (providerConfig) {
     const nextProviders = { ...providerConfig };
     for (const [providerId, provider] of Object.entries(providerConfig)) {
-      const models = provider.models;
+      const normalizedProvider = normalizeProviderSpecificConfig(providerId, provider);
+      const models = normalizedProvider.models;
       if (!Array.isArray(models) || models.length === 0) {
+        if (normalizedProvider !== provider) {
+          nextProviders[providerId] = normalizedProvider;
+          mutated = true;
+        }
         continue;
       }
-      const providerApi = resolveDefaultProviderApi(providerId, provider.api);
-      let nextProvider = provider;
-      if (providerApi && provider.api !== providerApi) {
+      const providerApi = normalizedProvider.api;
+      let nextProvider = normalizedProvider;
+      if (nextProvider !== provider) {
         mutated = true;
-        nextProvider = { ...nextProvider, api: providerApi };
       }
       let providerMutated = false;
       const nextModels = models.map((model) => {
@@ -434,105 +375,16 @@ export function applyLoggingDefaults(cfg: OpenClawConfig): OpenClawConfig {
 }
 
 export function applyContextPruningDefaults(cfg: OpenClawConfig): OpenClawConfig {
-  const defaults = cfg.agents?.defaults;
-  if (!defaults) {
-    return cfg;
-  }
-
-  const authMode = resolveAnthropicDefaultAuthMode(cfg);
-  if (!authMode) {
-    return cfg;
-  }
-
-  let mutated = false;
-  const nextDefaults = { ...defaults };
-  const contextPruning = defaults.contextPruning ?? {};
-  const heartbeat = defaults.heartbeat ?? {};
-
-  if (defaults.contextPruning?.mode === undefined) {
-    nextDefaults.contextPruning = {
-      ...contextPruning,
-      mode: "cache-ttl",
-      ttl: defaults.contextPruning?.ttl ?? "1h",
-    };
-    mutated = true;
-  }
-
-  if (defaults.heartbeat?.every === undefined) {
-    nextDefaults.heartbeat = {
-      ...heartbeat,
-      every: authMode === "oauth" ? "1h" : "30m",
-    };
-    mutated = true;
-  }
-
-  if (authMode === "api_key") {
-    const nextModels = defaults.models ? { ...defaults.models } : {};
-    let modelsMutated = false;
-    const isAnthropicCacheRetentionTarget = (
-      parsed: { provider: string; model: string } | null | undefined,
-    ): parsed is { provider: string; model: string } =>
-      Boolean(
-        parsed &&
-        (parsed.provider === "anthropic" ||
-          (parsed.provider === "amazon-bedrock" &&
-            parsed.model.toLowerCase().includes("anthropic.claude"))),
-      );
-
-    for (const [key, entry] of Object.entries(nextModels)) {
-      const parsed = parseModelRef(key, "anthropic");
-      if (!isAnthropicCacheRetentionTarget(parsed)) {
-        continue;
-      }
-      const current = entry ?? {};
-      const params = (current as { params?: Record<string, unknown> }).params ?? {};
-      if (typeof params.cacheRetention === "string") {
-        continue;
-      }
-      nextModels[key] = {
-        ...(current as Record<string, unknown>),
-        params: { ...params, cacheRetention: "short" },
-      };
-      modelsMutated = true;
-    }
-
-    const primary = resolvePrimaryModelRef(
-      resolveAgentModelPrimaryValue(defaults.model) ?? undefined,
-    );
-    if (primary) {
-      const parsedPrimary = parseModelRef(primary, "anthropic");
-      if (isAnthropicCacheRetentionTarget(parsedPrimary)) {
-        const key = `${parsedPrimary.provider}/${parsedPrimary.model}`;
-        const entry = nextModels[key];
-        const current = entry ?? {};
-        const params = (current as { params?: Record<string, unknown> }).params ?? {};
-        if (typeof params.cacheRetention !== "string") {
-          nextModels[key] = {
-            ...(current as Record<string, unknown>),
-            params: { ...params, cacheRetention: "short" },
-          };
-          modelsMutated = true;
-        }
-      }
-    }
-
-    if (modelsMutated) {
-      nextDefaults.models = nextModels;
-      mutated = true;
-    }
-  }
-
-  if (!mutated) {
-    return cfg;
-  }
-
-  return {
-    ...cfg,
-    agents: {
-      ...cfg.agents,
-      defaults: nextDefaults,
-    },
-  };
+  return (
+    applyProviderConfigDefaultsWithPlugin({
+      provider: "anthropic",
+      context: {
+        provider: "anthropic",
+        config: cfg,
+        env: process.env,
+      },
+    }) ?? cfg
+  );
 }
 
 export function applyCompactionDefaults(cfg: OpenClawConfig): OpenClawConfig {
diff --git a/src/plugin-sdk/plugin-entry.ts b/src/plugin-sdk/plugin-entry.ts
index 5f86d1ca98a..43be5b139f5 100644
--- a/src/plugin-sdk/plugin-entry.ts
+++ b/src/plugin-sdk/plugin-entry.ts
@@ -18,6 +18,7 @@ import type {
   ProviderAuthMethod,
   ProviderAuthMethodNonInteractiveContext,
   ProviderAuthResult,
+  ProviderApplyConfigDefaultsContext,
   ProviderBuildMissingAuthMessageContext,
   ProviderBuildUnknownModelHintContext,
   ProviderBuiltInModelSuppressionContext,
@@ -28,6 +29,7 @@ import type {
   ProviderDeferSyntheticProfileAuthContext,
   ProviderDefaultThinkingPolicyContext,
   ProviderDiscoveryContext,
+  ProviderFailoverErrorContext,
   ProviderFetchUsageSnapshotContext,
   ProviderModernModelPolicyContext,
   ProviderNormalizeConfigContext,
@@ -77,6 +79,7 @@ export type {
   ProviderCatalogResult,
   ProviderDeferSyntheticProfileAuthContext,
   ProviderAugmentModelCatalogContext,
+  ProviderApplyConfigDefaultsContext,
   ProviderBuiltInModelSuppressionContext,
   ProviderBuiltInModelSuppressionResult,
   ProviderBuildMissingAuthMessageContext,
@@ -84,6 +87,7 @@ export type {
   ProviderCacheTtlEligibilityContext,
   ProviderDefaultThinkingPolicyContext,
   ProviderFetchUsageSnapshotContext,
+  ProviderFailoverErrorContext,
   ProviderModernModelPolicyContext,
   ProviderNormalizeConfigContext,
   ProviderNormalizeToolSchemasContext,
diff --git a/src/plugins/provider-runtime.test.ts b/src/plugins/provider-runtime.test.ts
index 1b9ec00d63f..19801c9f07c 100644
--- a/src/plugins/provider-runtime.test.ts
+++ b/src/plugins/provider-runtime.test.ts
@@ -34,7 +34,10 @@ let buildProviderAuthDoctorHintWithPlugin: typeof import("./provider-runtime.js"
 let buildProviderMissingAuthMessageWithPlugin: typeof import("./provider-runtime.js").buildProviderMissingAuthMessageWithPlugin;
 let buildProviderUnknownModelHintWithPlugin: typeof import("./provider-runtime.js").buildProviderUnknownModelHintWithPlugin;
 let applyProviderNativeStreamingUsageCompatWithPlugin: typeof import("./provider-runtime.js").applyProviderNativeStreamingUsageCompatWithPlugin;
+let applyProviderConfigDefaultsWithPlugin: typeof import("./provider-runtime.js").applyProviderConfigDefaultsWithPlugin;
 let formatProviderAuthProfileApiKeyWithPlugin: typeof import("./provider-runtime.js").formatProviderAuthProfileApiKeyWithPlugin;
+let classifyProviderFailoverReasonWithPlugin: typeof import("./provider-runtime.js").classifyProviderFailoverReasonWithPlugin;
+let matchesProviderContextOverflowWithPlugin: typeof import("./provider-runtime.js").matchesProviderContextOverflowWithPlugin;
 let normalizeProviderConfigWithPlugin: typeof import("./provider-runtime.js").normalizeProviderConfigWithPlugin;
 let normalizeProviderModelIdWithPlugin: typeof import("./provider-runtime.js").normalizeProviderModelIdWithPlugin;
 let applyProviderResolvedModelCompatWithPlugins: typeof import("./provider-runtime.js").applyProviderResolvedModelCompatWithPlugins;
@@ -253,9 +256,12 @@ describe("provider-runtime", () => {
       buildProviderMissingAuthMessageWithPlugin,
       buildProviderUnknownModelHintWithPlugin,
       applyProviderNativeStreamingUsageCompatWithPlugin,
+      applyProviderConfigDefaultsWithPlugin,
       applyProviderResolvedModelCompatWithPlugins,
       applyProviderResolvedTransportWithPlugin,
+      classifyProviderFailoverReasonWithPlugin,
       formatProviderAuthProfileApiKeyWithPlugin,
+      matchesProviderContextOverflowWithPlugin,
       normalizeProviderConfigWithPlugin,
       normalizeProviderModelIdWithPlugin,
       normalizeProviderTransportWithPlugin,
@@ -383,6 +389,78 @@ describe("provider-runtime", () => {
     });
   });
 
+  it("resolves provider config defaults through owner plugins", () => {
+    resolveOwningPluginIdsForProviderMock.mockReturnValue(["anthropic"]);
+    resolvePluginProvidersMock.mockReturnValue([
+      {
+        id: "anthropic",
+        label: "Anthropic",
+        auth: [],
+        applyConfigDefaults: ({ config }) => ({
+          ...config,
+          agents: {
+            defaults: {
+              heartbeat: { every: "1h" },
+            },
+          },
+        }),
+      },
+    ]);
+
+    expect(
+      applyProviderConfigDefaultsWithPlugin({
+        provider: "anthropic",
+        context: {
+          provider: "anthropic",
+          env: {},
+          config: {},
+        },
+      }),
+    ).toMatchObject({
+      agents: {
+        defaults: {
+          heartbeat: {
+            every: "1h",
+          },
+        },
+      },
+    });
+  });
+
+  it("resolves failover classification through hook-only aliases", () => {
+    resolvePluginProvidersMock.mockReturnValue([
+      {
+        id: "openai",
+        label: "OpenAI",
+        hookAliases: ["azure-openai-responses"],
+        auth: [],
+        matchesContextOverflowError: ({ errorMessage }) =>
+          /\bcontent_filter\b.*\btoo long\b/i.test(errorMessage),
+        classifyFailoverReason: ({ errorMessage }) =>
+          /\bquota exceeded\b/i.test(errorMessage) ? "rate_limit" : undefined,
+      },
+    ]);
+
+    expect(
+      matchesProviderContextOverflowWithPlugin({
+        provider: "azure-openai-responses",
+        context: {
+          provider: "azure-openai-responses",
+          errorMessage: "content_filter prompt too long",
+        },
+      }),
+    ).toBe(true);
+    expect(
+      classifyProviderFailoverReasonWithPlugin({
+        provider: "azure-openai-responses",
+        context: {
+          provider: "azure-openai-responses",
+          errorMessage: "quota exceeded",
+        },
+      }),
+    ).toBe("rate_limit");
+  });
+
   it("resolves stream wrapper hooks through hook-only aliases without provider ownership", () => {
     const wrappedStreamFn = vi.fn();
     resolvePluginProvidersMock.mockReturnValue([
diff --git a/src/plugins/provider-runtime.ts b/src/plugins/provider-runtime.ts
index e51ed7432c6..77c5aa8bdcf 100644
--- a/src/plugins/provider-runtime.ts
+++ b/src/plugins/provider-runtime.ts
@@ -21,6 +21,7 @@ import type {
   ProviderCreateStreamFnContext,
   ProviderDefaultThinkingPolicyContext,
   ProviderFetchUsageSnapshotContext,
+  ProviderFailoverErrorContext,
   ProviderNormalizeToolSchemasContext,
   ProviderNormalizeConfigContext,
   ProviderNormalizeModelIdContext,
@@ -34,6 +35,7 @@ import type {
   ProviderPrepareExtraParamsContext,
   ProviderPrepareDynamicModelContext,
   ProviderPrepareRuntimeAuthContext,
+  ProviderApplyConfigDefaultsContext,
   ProviderResolveConfigApiKeyContext,
   ProviderSanitizeReplayHistoryContext,
   ProviderResolveUsageAuthContext,
@@ -593,6 +595,47 @@ export async function resolveProviderUsageSnapshotWithPlugin(params: {
   return await resolveProviderRuntimePlugin(params)?.fetchUsageSnapshot?.(params.context);
 }
 
+export function matchesProviderContextOverflowWithPlugin(params: {
+  provider?: string;
+  config?: OpenClawConfig;
+  workspaceDir?: string;
+  env?: NodeJS.ProcessEnv;
+  context: ProviderFailoverErrorContext;
+}): boolean {
+  const plugins = params.provider
+    ? [resolveProviderHookPlugin({ ...params, provider: params.provider })].filter(
+        (plugin): plugin is ProviderPlugin => Boolean(plugin),
+      )
+    : resolveProviderPluginsForHooks(params);
+  for (const plugin of plugins) {
+    if (plugin.matchesContextOverflowError?.(params.context)) {
+      return true;
+    }
+  }
+  return false;
+}
+
+export function classifyProviderFailoverReasonWithPlugin(params: {
+  provider?: string;
+  config?: OpenClawConfig;
+  workspaceDir?: string;
+  env?: NodeJS.ProcessEnv;
+  context: ProviderFailoverErrorContext;
+}) {
+  const plugins = params.provider
+    ? [resolveProviderHookPlugin({ ...params, provider: params.provider })].filter(
+        (plugin): plugin is ProviderPlugin => Boolean(plugin),
+      )
+    : resolveProviderPluginsForHooks(params);
+  for (const plugin of plugins) {
+    const reason = plugin.classifyFailoverReason?.(params.context);
+    if (reason) {
+      return reason;
+    }
+  }
+  return undefined;
+}
+
 export function formatProviderAuthProfileApiKeyWithPlugin(params: {
   provider: string;
   config?: OpenClawConfig;
@@ -663,6 +706,16 @@ export function resolveProviderDefaultThinkingLevel(params: {
   return resolveProviderRuntimePlugin(params)?.resolveDefaultThinkingLevel?.(params.context);
 }
 
+export function applyProviderConfigDefaultsWithPlugin(params: {
+  provider: string;
+  config?: OpenClawConfig;
+  workspaceDir?: string;
+  env?: NodeJS.ProcessEnv;
+  context: ProviderApplyConfigDefaultsContext;
+}) {
+  return resolveProviderRuntimePlugin(params)?.applyConfigDefaults?.(params.context) ?? undefined;
+}
+
 export function resolveProviderModernModelRef(params: {
   provider: string;
   config?: OpenClawConfig;
diff --git a/src/plugins/types.ts b/src/plugins/types.ts
index 8b760cd753b..2a67e341013 100644
--- a/src/plugins/types.ts
+++ b/src/plugins/types.ts
@@ -1,9 +1,9 @@
+import type { IncomingMessage, ServerResponse } from "node:http";
 import type { AgentMessage } from "@mariozechner/pi-agent-core";
 import type { StreamFn } from "@mariozechner/pi-agent-core";
 import type { Api, Model } from "@mariozechner/pi-ai";
 import type { ModelRegistry } from "@mariozechner/pi-coding-agent";
 import type { Command } from "commander";
-import type { IncomingMessage, ServerResponse } from "node:http";
 import type {
   ApiKeyCredential,
   AuthProfileCredential,
@@ -11,6 +11,7 @@ import type {
   AuthProfileStore,
 } from "../agents/auth-profiles/types.js";
 import type { ModelCatalogEntry } from "../agents/model-catalog.js";
+import type { FailoverReason } from "../agents/pi-embedded-helpers/types.js";
 import type { ProviderRequestTransportOverrides } from "../agents/provider-request-config.js";
 import type { AnyAgentTool } from "../agents/tools/common.js";
 import type { ThinkLevel } from "../auto-reply/thinking.js";
@@ -748,6 +749,30 @@ export type ProviderResolveWebSocketSessionPolicyContext = {
   sessionId?: string;
 };
 
+/**
+ * Provider-owned failover error classification input.
+ *
+ * Use this when provider-specific transport or API errors need classification
+ * hints that generic string matching cannot express safely.
+ */
+export type ProviderFailoverErrorContext = {
+  provider?: string;
+  modelId?: string;
+  errorMessage: string;
+};
+
+/**
+ * Provider-owned config-default application input.
+ *
+ * Use this when a provider needs to add global config defaults that depend on
+ * provider auth mode or provider-specific model families.
+ */
+export type ProviderApplyConfigDefaultsContext = {
+  provider: string;
+  config: OpenClawConfig;
+  env: NodeJS.ProcessEnv;
+};
+
 /**
  * Generic embedding provider shape returned by provider plugins.
  *
@@ -1288,6 +1313,20 @@ export type ProviderPlugin = {
   fetchUsageSnapshot?: (
     ctx: ProviderFetchUsageSnapshotContext,
   ) => Promise<ProviderUsageSnapshot | null | undefined> | ProviderUsageSnapshot | null | undefined;
+  /**
+   * Provider-owned failover context-overflow matcher.
+   *
+   * Return true when the provider recognizes the raw error as a context-window
+   * overflow shape that generic heuristics would miss.
+   */
+  matchesContextOverflowError?: (ctx: ProviderFailoverErrorContext) => boolean | undefined;
+  /**
+   * Provider-owned failover error classification.
+   *
+   * Return a failover reason when the provider recognizes a provider-specific
+   * raw error shape. Return undefined to fall back to generic classification.
+   */
+  classifyFailoverReason?: (ctx: ProviderFailoverErrorContext) => FailoverReason | null | undefined;
   /**
    * Provider-owned cache TTL eligibility.
    *
@@ -1359,6 +1398,15 @@ export type ProviderPlugin = {
   resolveDefaultThinkingLevel?: (
     ctx: ProviderDefaultThinkingPolicyContext,
   ) => "off" | "minimal" | "low" | "medium" | "high" | "xhigh" | "adaptive" | null | undefined;
+  /**
+   * Provider-owned global config defaults.
+   *
+   * Use this when config materialization needs provider-specific defaults that
+   * depend on auth mode, env, or provider model-family semantics.
+   */
+  applyConfigDefaults?: (
+    ctx: ProviderApplyConfigDefaultsContext,
+  ) => OpenClawConfig | null | undefined;
   /**
    * Provider-owned "modern model" matcher used by live profile/smoke filters.
    *