Le meilleur de l'écosystème IA et MCP, sélectionné chaque jour.
LangChain released two production-focused updates that make building and operating agents easier and more robust. The headline is Deep Agents v0.5, which introduces async subagents and richer multimodal filesystem support so agents can delegate long-running or parallel tasks without blocking and handle varied data types more reliably. Paired with that, LangChain added Arcade.dev’s catalog into LangSmith Fleet, giving agents access to 7,500+ agent-optimized tools through a single, secure gateway — simplifying discovery, authorization, and governance in production deployments.
Today's stories:
Taken together, these updates push LangChain toward more production-ready agent platforms: better concurrency and richer data handling on the runtime side, and a simpler, governed path to integrate many third-party tools.
Deep Agents v0.5 adds async subagents and expanded multimodal filesystem support, enabling non-blocking background delegation and richer data handling for complex agent workflows. These changes make it easier to build scalable, modular agent systems that run long-lived or parallel tasks.
LangChain announced Arcade's integration with LangSmith Fleet, giving agents access to Arcade's catalog of 7,500+ agent-optimized tools through a single secure gateway. This simplifies tool discovery and authorization for production agents, improving operational reliability and governance.
