AI Devtools Radar

LangChain

Agent frameworkswww.langchain.com11 published changes

2 sources under watch

Change timeline

featurehigh

Introduces new langchain.mcp namespace for adapting MCP servers into LangChain tools

Enables developers to integrate Model Context Protocol (MCP) servers as LangChain tools via MCPAdapter, supporting multiple server configurations and client-side response caching.

Evidence
+Third alpha of the `1.4.0` line. This release focuses on the new `langchain.mcp` namespace for adapting MCP servers into LangChain tools.

github.com

feature

LangSmith Engine released with improved issue detection capability

LangSmith users now have access to improved issue detection in LangSmith Engine, indicating a product enhancement for observability and debugging capabilities.

Evidence
+LangSmith New in LangSmith Engine: >2x better issue detection Trammell Saltzgaber Ben Tannyhill August 25, 2026

blog.langchain.dev · also seen at 1 more source

feature

OpenRouter integration now preserves provider information in response metadata

Users can now identify which provider fulfilled their request when using OpenRouter's multi-provider routing.

Evidence
+feat(openrouter): preserve provider in response metadata

github.com · also seen at 1 more source

positioning

Blog content focus shifts from agent comparison to managed agent solutions

LangChain is moving from positioning Deep Agents vs LangChain as competing approaches to promoting Managed Deep Agents as a LangSmith feature, indicating a product integration strategy.

Evidence
−Deep Agents vs LangChain vs LangGraph
+Managed Deep Agents is now in Public Beta

blog.langchain.dev

featurehigh

LangSmith introduces Managed Deep Agents in Public Beta

LangChain users can now use managed Deep Agents through LangSmith, expanding the platform's capabilities for agent orchestration.

Evidence
+LangSmith Managed Deep Agents is now in Public Beta Victor Moreira August 7, 2026 9 min

blog.langchain.dev · also seen at 1 more source

api

Fix tool_choice preservation in Anthropic integration

Tool selection behavior is now correctly preserved when passed by callers, ensuring expected tool invocation patterns.

Evidence
+fix(anthropic): preserve caller `tool_choice`

github.com

other

Featured blog articles updated with new content focus on voice agents and production CX agents

Blog readers see different featured content: shift from deep agents framework to voice agent evaluation and production CX agent case studies.

Evidence
−How Stripe Built their Knowledge AI Platform: A Company-Wide AI Agent on Deep Agents, Live in 1 Week Sofia Sulikowski August 3, 2026 10 How Stripe Built their Knowledge AI Platform: A Company-Wide AI Agent on Deep Agents, Live in 1 Week Conceptual Guide LangSmith Building Governed Agents: A Framework for Cost, Control, and Compliance Martha Janicki July 20, 2026 15 min Partner Proving the ROI of agentic AI in financial services Karan Singh David Tepper July 17, 2026 14 min
+No items found. How to evaluate voice agents: execution, outcomes, and experience Caroline di Vittorio August 4, 2026 11 No items found. How to evaluate voice agents: execution, outcomes, and experience Caroline di Vittorio August 4, 2026 11 min Conceptual Guide Customer Experience (CX) Agents in Production: Lessons from Lyft, Vodafone, and LATAM Airlines Jess Ou August 4, 2026 24 min How Stripe Built Kai, its Company-Wide AI Agent, on Deep Agents

blog.langchain.dev · also seen at 1 more source

featurehigh

New LangSmith LLM Gateway feature for runtime controls in production agents

Users can now apply runtime controls to production agents using the LangSmith LLM Gateway, expanding LangSmith's capabilities for agent management and safety.

Evidence
+LangSmith LLM Gateway: runtime controls for production agents

blog.langchain.dev · also seen at 1 more source

other

Released langchain-xai version 1.3.0

Users can now upgrade to version 1.3.0 which includes new features, bug fixes, and dependency updates.

Evidence
+langchain-xai==1.3.0

github.com · also seen at 1 more source

api

Dropped unsupported `stop` parameter from XAI integration

Code using the `stop` parameter with XAI models will need to be updated as this parameter is no longer supported.

Evidence
+fix(xai): drop unsupported `stop` parameter (#38335)

github.com

feature

New case study published on Apollo's Deep Agents implementation for GTM automation

Users can now learn from a real-world case study showing how Apollo rebuilt its AI assistant using Deep Agents to power full go-to-market loops.

Evidence
−No items found.
+Case Studies How Apollo Rebuilt Its AI Assistant on Deep Agents to Power the Full GTM Loop Sofia Sulikowski July 21, 2026 6 min

blog.langchain.dev · also seen at 1 more source