October 2025 Summaries

In September 2025, the GraphQL specification was updated to officially support operation descriptions, allowing queries, mutations, subscriptions, and fragments to have descriptions using the triple-quoted string syntax. This change resolved a documentation challenge faced by teams using the Cosmo MCP Gateway, which transforms trusted GraphQL operations into MCP tools for AI agents. Previously, teams had to rely on non-standard methods such as comments or custom directives to describe these operations. With the update, operation descriptions became standardized and portable, enhancing self-documentation without the need for additional configuration or custom parsing. This advancement simplifies the development of backend-for-frontends, improves AI tool selection, and reduces maintenance complexity across all GraphQL tools. The support for operation descriptions is available in Cosmo Router v0.262.0 and later, seamlessly integrating with WunderGraph's MCP Gateway to ensure consistent documentation and integration with AI systems.

From 2024 to 2025, AI security breaches like those involving Amazon Q, Vanna.AI, and EchoLeak highlighted the inadequacy of security controls designed for human users when applied to large language models (LLMs), which lacked runtime boundaries to prevent unauthorized code execution and data access. WunderGraph Cosmo proposes a solution by applying federation principles, such as persisted operations, scoped access, and signed configurations, to enforce runtime boundaries and prevent unverified execution. Case studies, including Vanna.AI's remote code execution vulnerability and EchoLeak's data leak via Copilot, demonstrate how trust misplaced in model output led to compromised environments. By implementing federation, systems can ensure that only approved actions are executed, credentials are tied to least privilege access, and unverified artifacts are blocked, transforming prompt injection from a breach into a blocked request. This governance framework prioritizes predictability over perfection, ensuring that AI systems operate within defined constraints and emphasizing proactive containment rather than reactive patching.

Connecting APIs to Large Language Models (LLMs) through the Model Context Protocol (MCP) is more efficient and secure when using a GraphQL facade with Persisted Operations (POs) rather than raw REST endpoints. REST APIs, while beneficial for developers and internal services, pose challenges for LLMs due to their granular and sprawling nature, leading to inefficiencies, context overload, and security risks. GraphQL, on the other hand, offers a unified schema and task-oriented approach, addressing these issues by combining microservices into a single, machine-readable schema and allowing precise data fetching. Persisted Operations enhance security and predictability by allowing only predefined queries and mutations, ensuring task-level granularity, controlled context, and governance, which are crucial for stable tool usage and consistent operation contracts. This approach abstracts complexity and provides a stable, predictable interface, making it particularly effective for connecting APIs to LLMs via MCP.

GraphQL and CQRS (Command Query Responsibility Segregation) naturally align, where GraphQL queries correspond to the read operations and mutations to the write operations in CQRS, allowing for optimized data fetching and modification with clear contracts. This architecture benefits from GraphQL Federation, enabling distributed teams to work autonomously on different parts of the API while maintaining cohesion through subgraphs and shared entities. Event-Driven Federated Subscriptions (EDFS) further enhance this setup by using minimal events published to brokers like NATS, allowing the GraphQL layer to enrich these events with additional data seamlessly. The Cosmo Router facilitates this process by managing subscriptions, enriching events, and scaling efficiently, which simplifies real-time event consumption for frontend developers. Additionally, persisted operations in Cosmo enhance security by only allowing pre-registered queries, reducing the API's attack surface and enabling fine-grained control over client operations.

GraphQL Federation has evolved from its initial version, V1, which was rigid and inflexible, to a more developer-friendly V2, introducing simpler syntax and greater flexibility. V1, released publicly in 2019, introduced the concept of entities, which are GraphQL Object types defined with a @key directive, allowing for resolving objects over multiple services. V2, announced in late 2021, removes the concept of "origin entities" and the need for special syntax for entity definitions across subgraphs, promoting easier migration and compatibility. Additionally, V2 introduces stricter rules for using the @external directive, ensuring fields and types conform to specific paths using @provides or @requires directives. The text also discusses a conceptual "Version 1.5," which represents subgraphs that are interpreted as V1 by V2 routers but are not fully compatible with true V1 composition, reflecting a hybridization of the two versions. This evolution aims to enhance cooperation across teams and safeguard against unintended changes in subgraphs, with V2 supporting backwards compatibility but sometimes requiring modifications for successful composition with V1.