diff --git a/docs/auto_router/feature_history.md b/docs/auto_router/feature_history.md
index eb5c2ad21..094ed8317 100644
--- a/docs/auto_router/feature_history.md
+++ b/docs/auto_router/feature_history.md
@@ -17,8 +17,10 @@ Merged after the v1.100.0 release candidate was cut. Available in `v1.101.0-dev`
 - **Shadow evals on teams and users.** Target a `key`, `team`, or `user`, so JWT-authenticated traffic can be evaluated. [#39015](https://github.com/BerriAI/litellm/pull/39015).
 - **Shadow evals across several routers.** Compare multiple router configs on one job's sampled traffic, paired. [#39028](https://github.com/BerriAI/litellm/pull/39028).
 - **1M context preset.** [#39490](https://github.com/BerriAI/litellm/pull/39490).
+- **Mid-task stall escalation.** `stall_escalation_enabled: true` reads the assistant's own recent tool calls and bumps a request one tier when it's stuck in a retry loop, the same ladder `escalation_keywords` uses. Off by default. [#39809](https://github.com/BerriAI/litellm/pull/39809). [Post](/blog/auto-router-stall-escalation).
+- **One-click Auto Router setup.** Configure automatically checks the chat model groups your proxy already serves and fills all four tiers, mixing providers when needed, without picking a template first. [#39693](https://github.com/BerriAI/litellm/pull/39693).
 
-Posts: [Route on Context Size and Modality](/blog/auto-router-more-routing-configurations).
+Posts: [Route on Context Size and Modality](/blog/auto-router-more-routing-configurations), [Mid-Task Stall Escalation](/blog/auto-router-stall-escalation).
 
 ## v1.100.0 (release candidate)
 
diff --git a/docs/auto_router/index.md b/docs/auto_router/index.md
index 06b84cd89..e3cf0bcc8 100644
--- a/docs/auto_router/index.md
+++ b/docs/auto_router/index.md
@@ -21,11 +21,12 @@ Already testing it? Share your results in [discussion #32168](https://github.com
 
 <AutoRouterDiagram />
 
+- **One-click setup.** Configure automatically checks the models your proxy already serves and fills all four tiers for you; no template to pick.
 - **One model name in your clients.** The gateway classifies each request and picks the model.
 - **Any model, any provider, per tier.** A single model, a random pool, or a Thompson-sampled pool.
 - **Three classifiers.** Sub-millisecond heuristic scorer, a small LLM, or keyword rules.
 - **Savings reported per request.** Against an all-frontier baseline, in the logs and in Cost Optimization.
-- **Agent-safe.** Prompt caching, context-window escalation, modality routing, and optional session pinning.
+- **Agent-safe.** Prompt caching, context-window escalation, modality routing, mid-task stall escalation, and optional session pinning.
 
 ## Results
 
@@ -123,6 +124,7 @@ items={[
 
 ## Release posts
 
+- [Mid-Task Stall Escalation](/blog/auto-router-stall-escalation): bumps a request one tier when it's stuck in a retry loop
 - [Auto Router v2](/blog/autorouter-v2): one router for complexity, semantic, and adaptive routing
 - [1-click presets and Test Routing](/blog/auto-router-setup-and-testing)
 - [Savings tab and per-request classifier cost](/blog/auto-router-spend-visibility)