MCP News

New Relic has announced support for the Model Context Protocol (MCP) within its observability platform. * This integration allows developers and operations teams to gain deeper insights into AI applications, particularly those leveraging Large Language Models (LLMs). * The platform can now ingest data generated by MCP, providing a comprehensive view of AI assistant interactions and associated infrastructure. * New Relic's enhanced capabilities enable monitoring of token usage, model performance, and costs for AI applications, including those built with Anthropic's Claude. * The support for MCP aims to bridge the gap between AI development and production, offering better debugging and optimization tools for AI-driven systems.

Sentry has launched its own Model Context Protocol (MCP) server, designed to integrate with AI assistants like Claude on desktop. * The Sentry MCP server allows AI models to query an organization's Sentry data, enabling them to retrieve recent errors, performance issues, and detailed event information. * It leverages Sentry's public API to fetch real-time data, providing AI with up-to-date operational context for debugging and analysis. * Users can run the MCP server locally via `sentry-cli` or deploy it remotely, configuring their AI assistants to connect to it. * This integration aims to enhance AI's ability to act as a 'copilot' for developers, providing relevant Sentry insights directly within AI-driven workflows.

GitHub has announced the public preview availability of its remote Model Context Protocol (MCP) Server. This new server enables AI assistants and a wide range of AI tooling to directly access GitHub-hosted code, issues, pull requests, and other repository data through the MCP. The service is designed to facilitate deeper, more accurate context sharing for AI agents and workflows operating on GitHub resources. This development aims to significantly improve the accuracy, relevance, and overall performance of AI-powered developer tools within the broader developer ecosystem. The remote MCP Server is now accessible for all GitHub users to explore and integrate into their AI-driven processes, with ongoing enhancements and additional capabilities anticipated.

JetBrains WebStorm announced a livestream event focused on building a Model Context Protocol (MCP) server. The session, led by Ekaterina Petrova and Ekaterina Ryabtseva, demonstrated how to create an MCP server from scratch. Attendees learned to implement basic functionality within an existing Express.js application, including handling context retrieval and action execution requests. Key details: * The livestream was titled 'Building an MCP Server: Live Coding in WebStorm'. * It covered implementing MCP server capabilities within an Express.js application. * Participants learned how to expose data and actions to AI models. * The session showcased practical steps for setting up and debugging an MCP server in WebStorm.

The Model Context Protocol (MCP) is presented as a significant advancement for integrating AI more effectively into complex marketing workflows. MCP allows AI models to process and act upon rich, real-time contextual information, moving beyond the limitations of single prompt windows. * The protocol enhances AI's ability to access and utilize external data sources, tools, and internal enterprise systems. * It is critical for developing sophisticated AI agents capable of managing multi-step processes and dynamic data environments. * MCP supports the creation of AI assistants that can autonomously interact with various systems, improving aspects like content generation, personalization, and campaign management. * The article highlights MCP's role in enabling AI to deliver more precise and relevant outputs by providing a deeper understanding of the operational context.

New Relic has announced support for the Model Context Protocol (MCP), aiming to provide true end-to-end observability for AI applications. * The integration allows developers to trace user requests through large language models (LLMs) and their various integrations, offering unprecedented visibility into the AI application stack. * New Relic's platform will now ingest and process MCP data, enabling comprehensive monitoring of AI workflows from user input to LLM response and tool usage. * This support is crucial for debugging, performance optimization, and understanding the behavior of complex AI systems, especially those using tools and RAG. * It helps address the 'black box' challenge of LLMs by providing transparent insights into their operations and interactions within an application.

Varonis has announced its data security platform is the first to support the Model Context Protocol (MCP). * This integration enables AI assistants, such as Anthropic's Claude, to securely access and utilize sensitive enterprise data as context. * Varonis ensures that only authorized, non-sensitive information is provided to AI models via MCP, maintaining data privacy and compliance. * The new capability allows AI assistants to offer more accurate and relevant responses by incorporating secure enterprise context into their operations. * The development addresses the challenge of securely leveraging internal data for AI without exposing it or risking data leakage.

Palo Alto Networks has launched a new Model Context Protocol (MCP) Server as part of its Prisma AI Security (Prisma AIR) platform. * The MCP Server is designed to secure AI agent innovation by providing a controlled and monitored pathway for data context. * It helps organizations enforce data governance and prevent sensitive information from being inappropriately shared with large language models. * This offering positions Prisma AIR as a critical component for enterprises looking to safely deploy and manage AI agents at scale. * The solution aims to address the security challenges inherent in providing AI agents access to enterprise data.

The Pipedream platform facilitates integrating OpenAI's capabilities within the Model Context Protocol (MCP) ecosystem. It enables developers to construct MCP servers through Pipedream workflows, utilizing OpenAI's Node.js SDK for implementing custom tool logic. These servers expose `fetch` and `execute` endpoints, providing structured tools and data directly to AI assistants compatible with MCP, notably Claude 3 Opus. The documentation includes code examples for creating tool definitions, processing tool calls (including handling `x-anthropic-tool-use-id`), and delivering contextually relevant responses back to the AI model.

A new attack vector, "Output Poisoning," targets Model Context Protocol (MCP) servers and the broader AI assistant ecosystem. * The attack involves injecting invisible, zero-width characters into Large Language Model (LLM) outputs that appear benign to users. * MCP servers can transmit these unsanitized characters, allowing them to bypass security filters and alter the behavior of downstream AI assistant tools and systems. * This can lead to severe consequences such as command injection, data exfiltration, or unauthorized execution within environments processing the "poisoned" output. * The research demonstrates how this vulnerability allows attackers to compromise systems even when outputs seem clean, highlighting a significant security concern for MCP server operators and AI assistant developers.

Data platforms are critical for the success of AI agents and the Model Context Protocol (MCP), enabling AI to access and process high-quality, real-time contextual information. * AI agents require extensive context from various data sources, including enterprise applications, internal systems, and external APIs, to perform complex tasks effectively. * MCP relies on robust data infrastructure to provide structured, relevant, and continuously updated context to large language models (LLMs), ensuring more accurate and reliable responses. * Key components of a supportive data platform include data integration, vector databases for semantic search, real-time data pipelines, and strong data governance. * Investing in a comprehensive data platform is essential for organizations to scale AI agent deployments, improve model performance, and achieve the full potential of AI-driven automation.