May 2025 Summaries

Firebase Studio offers features beyond its core prototyping, code editing, and App Hosting capabilities to speed up app development and iteration. Users can communicate design changes to the App Prototyping agent by annotating the live preview, uploading wireframes or reference images, and selecting individual interface elements before describing requested modifications. Previous app states can be restored when changes introduce unwanted results or bugs, while Code view allows developers to choose among available Gemini models for coding assistance. Collaboration options include sharing an entire workspace with trusted Google-account users, though shared access can expose sensitive files and simultaneous edits may overwrite changes, as well as making a live preview public through a QR code or URL for testing on other devices.

Firebase Studio, Google’s cloud-based AI workspace for building full-stack applications from prompts or code, has introduced updates aimed at faster prototyping and richer app development. A new Builder.io integration lets users export Figma designs into Firebase Studio and use Gemini to extend interfaces, add screens, and connect backend services, while the platform can now detect needs for authentication or databases and recommend and provision Firebase Auth and Cloud Firestore during deployment. Other additions include Unsplash image replacement, native Gemini-powered image generation within apps, improved mobile prototyping, and shareable prompt URLs that do not expose an entire workspace. Firebase Studio has also incorporated newer Gemini 2.5 Flash and Pro capabilities, Gemini API features such as audio output and Live API, and support for experimenting with Gemma models, positioning the platform as a central environment for creating, deploying, and operating AI-enabled applications.

Firebase has renamed Vertex AI in Firebase to Firebase AI Logic, Vertex AI Gemini API to Agent Platform Gemini API, and Vertex AI to Agent Platform, positioning Firebase AI Logic as a way to add generative AI to mobile, web, and game applications through direct client-side SDK access or the server-side Genkit framework. New client features include preview access to the Gemini Developer API alongside the enterprise-oriented Agent Platform Gemini API, AI monitoring dashboards, Unity support, Gemini-based image generation and editing, and experimental hybrid on-device and cloud inference using Gemini Nano in Chrome. Genkit, the open-source server-side framework, supports JavaScript, Python, and Go and provides unified access to models from Google and other providers, along with tools for structured outputs, retrieval-augmented generation, agents, testing, tracing, and deployment. Updates also enable dynamic model lookup in Genkit for JavaScript so developers can access newer Gemini models without package updates, while expanded Developer UI features simplify model exploration and use of model-specific capabilities.

Firebase has introduced an experimental Model Context Protocol server within the Firebase CLI, enabling AI coding assistants and IDE clients to interact with Firebase tools and services through a standardized local integration. The server offers more than 30 project-aware tools for tasks including initializing Firebase projects, managing Authentication users, querying and modifying Firestore data, validating security rules, sending Cloud Messaging notifications, handling Remote Config, reviewing production crashes, working with Cloud Storage URLs, and executing Data Connect GraphQL queries. It can be configured through an MCP client to run the latest Firebase CLI version, automatically detect a project’s firebase.json configuration, and activate relevant tools, while optional directory and feature filters provide additional control. Users generally must authenticate with the Firebase CLI before using protected capabilities. Firebase describes the release as early-stage and expects potentially significant changes, with planned additions including broader product coverage, documentation integration, deployment support, and Firebase Studio integration, while inviting community feedback through GitHub.

Firebase App Hosting, recently released for general availability, has expanded beyond GitHub-based deployments to support direct deployments through Firebase Studio, the Firebase CLI, and Terraform. Firebase Studio enables deployment from its workspace and includes onboarding and monitoring tools, while the CLI can upload local source code and configuration alongside other Firebase resources, building applications through Cloud Build before deploying them to Cloud Run and Cloud CDN. Terraform supports infrastructure-as-code workflows and allows developers to deploy pre-built container images, offering greater control over builds, dependencies, and rollout speed. App Hosting has also improved local development through automatic Firebase Emulator Suite configuration: JavaScript apps using argument-free Firebase initialization connect to emulated services locally rather than production resources, helping developers test safely without affecting live data.

Firebase has released a Preview of Firebase AI Logic for Unity, a client-side SDK that lets Unity developers integrate Gemini generative AI directly into games and Android XR applications without complex backend infrastructure. The SDK supports the Gemini Developer API, which offers a no-cost prototyping tier and pay-as-you-go production option, as well as the enterprise-focused Vertex AI Gemini API, now renamed Agent Platform Gemini API, with availability and governance features. It enables multimodal inputs and outputs, bidirectional audio streaming, image generation, and context-aware conversational NPCs, supporting applications such as interactive storytelling, personalized game environments, and XR artwork that responds to users’ surroundings. Built with production considerations, it integrates with Firebase Authentication, App Check, and Cloud Storage, supports Android XR Developer Preview 2 in Unity 6, and is compatible with Unity 2021 LTS and later on Android and iOS.

Google announced a series of Firebase updates at I/O aimed at supporting AI-enabled application development across the full lifecycle, alongside revised product names effective August 2026, including Firebase AI Logic and Agent Platform. Firebase Studio, a cloud-based AI workspace powered by Gemini 2.5, adds Figma imports through Builder.io, image replacement using Unsplash, and automated recommendations and provisioning for Firebase backends such as Authentication, Firestore, and App Hosting. Firebase AI Logic expands client-side access to Gemini and Imagen models through both Gemini Developer and Agent Platform Gemini APIs, introduces hybrid local and cloud inference with Gemini Nano on Chrome, previews Unity support for games and Android XR, enables Gemini-based image generation and editing, and enhances AI monitoring. Server-side AI development remains supported through the open-source Genkit framework, while a preview Firebase MCP Server in the Firebase CLI enables AI assistants to inspect and manage Firebase projects, app data, crashes, and configurations. Other releases include local CLI and Terraform deployment options for App Hosting, automatic use of emulated services during local testing, AI-assisted onboarding and schema generation for Data Connect with Cloud SQL for PostgreSQL, and a forthcoming carrier-based phone-number verification capability intended to simplify onboarding and reduce SMS-related costs and abuse.

At Google I/O 2025, Firebase announced new capabilities for Data Connect, its Cloud SQL for PostgreSQL backend service, which has since been renamed SQL Connect. Expanded transaction support allows operations within an atomic transaction to use values generated or returned by earlier steps, including dynamic expressions for timestamps, authentication tokens, and computed data. Gemini-powered tools in the Firebase Console and VS Code through Gemini Code Assist can help developers generate and refine schemas, GraphQL queries, and mutations, then deploy configurations. The update also introduces server rules for Model Context Protocol servers to enable more controlled AI-agent access to Firebase-managed data, along with Firebase Admin SDK bulk import functions such as insertMany and upsertMany for data seeding, migrations, and batch updates.

Firebase Studio has upgraded its built-in App Prototyping agent to Gemini 2.5 for all preview users, following the creation of more than 1.5 million workspaces across applications ranging from e-commerce and CRM tools to games and educational services. Google says the model’s stronger reasoning and coding capabilities produce more modern user interfaces and support more complex features, including AI workflows, calculations, sound, images, API integrations, and iterative design changes through natural-language prompts. Demonstrations included a music tutor that added instrument support, a scheduling tool integrating Google Tasks and Calendar APIs, interactive quantum-physics lessons, a social-media planning platform, a visual AI-workflow builder, an AI-generated workout planner, and a Mancala game with an AI opponent. Additional updates improve mobile-oriented UI support and provide developers with more detailed explanations of file changes, while Firebase Studio plans to add further Gemini models and capabilities based on user feedback.

Firebase Data Connect, now renamed SQL Connect, introduces atomic operations intended to maintain data consistency and reduce race conditions in collaborative or high-traffic applications. Numeric fields of types including Int, Int64, Float, Date, and Timestamp can be updated atomically with `inc` and `dec`, illustrated through banking balance changes. For non-Vector list fields, `prepend` and `append` insert items at the beginning or end, while `add` inserts an item only if it is not already present and `remove` deletes every matching occurrence. These operations are presented as tools for supporting common application features such as counters and to-do lists while making SQL Connect a more capable database service.

Firebase Data Connect, now renamed SQL Connect, uses GraphQL queries to retrieve schema-defined backend data for applications, with queries written for client use after deployment or run ad hoc from privileged tools such as the Firebase console, VS Code extension, and Admin SDK. A query includes an operation name, typed arguments, optional directives, source tables, filters, ordering, and selected return fields; directives include required client-access authentication policies through @auth, result validation through @check, and response omission through @redact. Local tooling automatically generates single-table fields for retrieving individual records or filtered lists, as well as relational fields for multi-table operations, and these can be explored through the VS Code extension, generated files, local documentation, and the Firebase console. Developers use generated fields to define controlled application queries that expose only selected fields, preventing clients from directly accessing database data or secret fields, while automatically generated SDKs translate deployed GraphQL queries into functions for application platforms.