MCP App StoreLe meilleur de l'écosystème IA et MCP, sélectionné chaque jour.
Sources
SafetyKit implements advanced risk-assessment agents using OpenAI's GPT-5 to automate content moderation and compliance. This demonstrates the scaling of agentic safety systems for high-accuracy enterprise moderation.
ModernBERT now supports multiple languages through mmBERT, bringing the efficiency and performance of the ModernBERT architecture to multilingual NLP tasks. This expansion allows for better cross-lingual representation and retrieval.
The MCP Registry launched in preview — an open catalog and API for discovering publicly available MCP servers. Provides a standardised way for clients to discover servers, check metadata, and verify compatibility without relying on community-maintained lists. The infrastructure layer that makes the MCP ecosystem more navigable as it grows.
New research from OpenAI explores the root causes of LLM hallucinations. The findings emphasize how refined evaluation techniques can significantly improve the reliability and honesty of AI outputs.
The official PHP SDK for MCP is now generally available, built in collaboration with the PHP Foundation and Symfony. Brings idiomatic PHP support to MCP server development, opening the protocol to the large PHP developer community. Part of MCP's push to have official SDKs across all major languages.
Google has released EmbeddingGemma, a highly efficient embedding model designed for high-performance retrieval and semantic search. It provides a strong balance of accuracy and speed for RAG pipelines and vector databases.
Hugging Face introduces ahead-of-time (AOT) compilation for ZeroGPU Spaces to improve performance. This optimization reduces startup times and increases execution speed for GPU-accelerated applications.
OpenAI releases a more advanced speech-to-speech model and updates to the Realtime API. Key additions include native MCP server support, image input capabilities, and SIP phone calling integration.
OpenAI and Anthropic conducted a first-of-its-kind joint safety evaluation of each other's models. The study focuses on misalignment, hallucinations, and jailbreaking, emphasizing the importance of cross-lab collaboration in AI safety.
Anthropic is piloting a native Claude integration inside Chrome, bringing AI assistance directly into the browser without switching tabs or apps. The pilot explores how Claude can help with web research, form-filling, summarisation, and contextual tasks as users browse. An early look at where browser-native AI assistance is heading.
A deep dive into how MCP handles OAuth 2.1 client registration — the mechanism that lets MCP clients and servers establish trust without a pre-existing relationship. Explains the two core challenges (dynamic registration, trust bootstrapping) and the design choices made to solve them in a way that scales to the open MCP ecosystem. Essential for developers building MCP clients or auth-enabled servers.
Integration of Claude with Hugging Face leveraging the Model Context Protocol (MCP) to enable image generation capabilities. This expands Claude's utility by connecting it to specialized HF models via a standardized interface.
A comprehensive technical guide on developing and scaling production-grade CUDA kernels. Essential reading for developers optimizing deep learning performance at the hardware level.
Explores the application of the Model Context Protocol (MCP) to connect AI agents with specialized research tools. Highlights how standardized interfaces can accelerate scientific discovery and data analysis.
Arm and ExecuTorch 0.7 focus on optimizing generative AI for edge devices. This release expands the capability to run efficient AI models on ARM hardware, bridging the gap between cloud and on-device AI.
Arm introduces Neural Super Sampling, utilizing AI to enhance image resolution and quality efficiently. This development highlights the push for on-device AI acceleration for visual tasks.
TextQuests evaluates the performance of LLMs in the complex environment of text-based video games. It examines the models' ability to maintain state and handle emergent gameplay mechanics.
Hugging Face introduces AI Sheets, a tool enabling developers to manipulate and analyze datasets using open AI models directly within a spreadsheet interface.
Technical guide on utilizing Accelerate ND-Parallel for efficient multi-GPU training. Provides essential patterns for scaling AI model training across multiple devices.
OpenAI has released GPT-5 to its API platform, featuring significantly improved reasoning capabilities and enhanced controls for developers. The model demonstrates best-in-class performance on complex real-world coding tasks.