Google announced the preview launch of Connected Apps Platform on October 28, allowing developers to connect applications and services to Search and Gemini. The platform brings registration, testing, review and integration management into one console. Access is opening in phases through applications to participate, and the company will add developers, markets and Google products over the coming months.
Connected apps allow users to consult information from their own accounts and initiate actions in a service during an AI conversation. Developers decide which functions to expose and which eligible products they want their integration to appear in. Before publication, the app undergoes Google's review; configuration and tester access come before public availability.
A shared integration for web services and mobile applications
Google's recommended route for capabilities that need to operate across platforms is a cloud-hosted MCP server. Model Context Protocol standardizes the model's access to a service's authorized tools and data. A business that already has web APIs or an MCP server can use them as the starting point for its integration, following the platform's specifications.
The server exposes functions for eligible experiences on the web, Android and iOS. A second route, Android MCP, uses AppFunctions and its Jetpack library for functions executed locally on a phone. It is intended to integrate Android applications with agents and assistants such as Gemini and can use context available on the device.
Each registered app uses a single connection type, selected when it is created. That type cannot subsequently be changed for the same registration. The app's category defines baseline functional requirements and also remains fixed once selected. These conditions define the structure in which developers register their services in the new console.
Authorization, business identity and version control
Google Connected Console manages an app's public details, MCP connection, test users and publication. Server integration requires a publicly accessible HTTPS endpoint and verification of domain ownership. Within an organization, the account administrator or owner accepts the program agreement before invited members can continue configuring the integration.
Authentication can cover communication between services or individual user authorization, depending on the app's capabilities. In the latter case, the person links their account with that service to their Google Account. The platform's documentation specifies OAuth for delegated access, so the service performs operations within the permissions granted.
The public identity includes the app's name, icon, description, capabilities and example requests. Google requires at least three useful tasks beyond the category's baseline capabilities and three examples that help users understand what the service can do. Public descriptions explain functionality, while technical instructions for the model are managed separately in the server integration.
The console takes a snapshot of the tool definitions from the MCP server. During preview, developers can synchronize changes and test them. After publication, updating tools, descriptions or server details requires creating a new version of the connected app, maintaining control over the configuration offered to users.
Unpublished versions are visible only to accounts added by the developer to the tester list, for which the console requires Gmail addresses. When a version is submitted for review, its configuration is locked and the console creates a separate editable draft. The team can therefore prepare subsequent changes while Google checks the submitted version.
Testing is also supported by Gemini Apps Developer Mode, which allows inspection of event logs, tool definitions and authentication. The environment includes date, time and location simulation to check behavior. These tools are intended to evaluate the integration before approval is requested and the app is published.
Countries, languages and actions are configured explicitly
Developers select the countries and regions, languages and Google products in which a version can be available. Tools can be enabled individually, and an explicit confirmation option covers operations requiring the user's approval. MCP definitions marking an action as destructive automatically preselect this confirmation requirement.
Access to additional signals, such as location or the account's language, must be justified by the functionality provided. Platform policy requires the minimum necessary permissions, disclosure of data collection and use, and a public privacy policy. Data and credentials must be protected at rest and in transit, and users must be able to manage and delete their data from the service.
Apps are also assessed for stability and for consistency between their descriptions and actual behavior. Platform policy requires integrations without advertising content; their own brand elements, such as the name and icon, are permitted. After launch, developers can monitor traffic, latency and error rates, indicators Google includes in service management.
What is available in Search and Gemini
Google says it has launched more than 50 connected apps across Search and Gemini combined, in areas such as productivity, shopping and ticketing. Integrations with Adobe, Experian, Localiza and Peloton are available in Gemini, with their expansion to Search planned for the coming weeks. An integration with Calm has been announced for both products.
For Search users, the existing feature is available in AI Mode in the United States and in English. It requires sign-in with a Google Account and, for many services, linking an account with the app. The list includes Canva for design and Instacart for shopping. Completing certain operations, such as paying for an order or modifying a design, continues in the partner's application.
In Gemini, conditions are set separately for each app. For example, the Adobe integration is intended for personal Google Accounts, users aged at least 18 and requests in English, in countries where both Gemini and the partner's service are available. The new developer platform retains this differentiated access management as participation in the preview expands in phases.
