September 2026 Summaries

Apollo Mock is a new tool for Kotlin applications that generates schema-compliant mock GraphQL data with an LLM, reducing the maintenance burden of manually created fixtures, faker setups, and hardcoded JSON. Configured through a Gradle plugin, it reads a GraphQL schema and produces persistent JSON entity files, using a local Ollama model by default or Anthropic’s API for more complex schemas. Because the LLM is used only during generation, the resulting data can be served quickly and deterministically without ongoing model access or API keys, while schema updates preserve existing entity IDs and values where possible. Developers can add Apollo Mock’s server module to expose the generated data through a local GraphQL endpoint and interactive sandbox, allowing ApolloClient or other GraphQL clients to interact with it like a real backend. Apollo Mock is in its first release, also offers direct CLI use outside Kotlin, and invites feedback through its documentation, GitHub, and Discord.

Apollo announced that its planned addition of second IP addresses for covered Uplink API services is now active, requiring organizations that restrict outbound traffic by IP address to allowlist both the existing and new addresses for uplink.api.apollographql.com and persisted-queries.api.apollographql.com. Organizations that have not added the new addresses may experience Router or Gateway failures when polling Uplink for schemas or configuration, while those using domain-based allowlisting or no outbound restrictions need take no action. Apollo identifies its API IP Allowlist Policy page as the authoritative source for current and future changes, with at least 60 days’ notice for additions, and directs users with questions to Technical Support.

Apollo Client 4.3 introduces type-safety and cache-handling improvements, led by native custom scalar support, the project’s most requested feature, which lets developers define parsing, serialization, and type-checking logic once and use native values such as Date objects in query results, variables, and cache writes. The release also enables applications to declare a specific cache implementation type through TypeScript overrides, removing repeated type casts, and improves GraphQL Codegen support for incremental queries by narrowing deferred data types once data is complete. Supporting custom scalars required broader changes to how Apollo Client handles incremental @defer and @stream responses: new defer-aware cache reads and data pruning ensure transformed cache values are returned consistently while fields within incomplete incremental boundaries remain unavailable until delivered by the network. Apollo Client also redesigns cache-feud detection, always writing network results to the cache and tracking repeated missing fields to prevent infinite refetch cycles while retaining complete transformed values; it now issues warnings when this protection activates. Other updates include a minimum TypeScript version of 5.9, skipToken support in useSubscription, and stricter configuration options for the from field-policy setting.

Apollo Client 4.3 introduces native custom scalar support, enabling developers to centrally define parsing, serialization, and type-detection logic for values such as dates and URLs, then use parsed types consistently in cache reads, writes, query variables, and complex input objects. The release also adds TypeScript improvements, including configurable cache typing through TypeOverrides and improved GraphQL Codegen types that narrow incremental query data more accurately when streamed or deferred results become complete. Supporting custom scalars required broader changes to incremental delivery behavior: Apollo Client now uses defer-aware cache reads and data pruning to return transformed cache values without exposing deferred fields before the network has delivered them. Cache feud detection was also redesigned so results are always written to the cache while repeated missing-field patterns prevent infinite refetches, improving consistency and addressing a longstanding refetch issue. Additional changes include a new custom-scalars code generation plugin, support for skipToken in useSubscription, stricter typing options for the from setting, and a minimum supported TypeScript version of 5.9.

Apollo GraphOS and Apollo MCP Server received 2026 API Awards at API World in Santa Clara, with GraphOS winning Best Innovation in API Orchestration and MCP Server winning Best Innovation in AI API. The independent DevNetwork API Advisory Board selected winners based on technical innovation, industry impact, community adoption, and advancement of API practices. Apollo GraphOS is presented as a platform for federating and governing organizational APIs, while Apollo MCP Server uses the Model Context Protocol to let AI agents securely call existing GraphQL APIs and access governed production data. Apollo will showcase its agentic AI and API strategy at Apollo Summit 2026 in San Francisco from October 6–8, featuring speakers and enterprise sessions from organizations including American Airlines, Expedia Group, and Yahoo.