# What is Connector? Also called app connector. A connector is a prebuilt, reusable component that links a platform to one specific external service, packaging the authentication, endpoints, and data mapping that an integration to that service requires. Users typically enable one and sign in rather than writing code. The term is vendor specific, and the depth of what a connector actually does varies widely between products. A typical connector bundles a credential flow, a set of supported operations, a mapping between the external service's data model and the platform's own, and some handling for pagination and rate limits. Platforms maintain catalogs of them so that adding a new service becomes a configuration step for the user instead of an engineering project for the vendor. In AI products, connectors are how an assistant gains access to a calendar, a code host, a customer record system, or a document store. Standardized tool protocols are changing how they are built, because a single server implementation can serve many clients at once rather than each platform maintaining its own private adapter for the same external service. Do not assume a connector's name tells you its scope. Two products can both list the same service while one reads a single object type and the other supports full two way synchronization. Permissions are the other blind spot, since a connector often requests broad scopes at setup, so what an agent can reach later is decided in that moment by whoever clicked approve. Connectors reduce work but concentrate dependency. When an external service changes its interface, everyone using that connector waits for the maintainer. A connector that silently stops synchronizing is worse than no connector at all, because people keep trusting stale data, so monitoring whether a connection is still alive belongs with the connector rather than with the user noticing. ## Key points - Prebuilt adapter linking a platform to one external service. - Bundles authentication, supported operations, and data mapping. - Enabled through configuration rather than custom code. - Depth varies enormously between products using the same name. - Scopes approved at setup decide what an agent can reach later. ## In practice A workspace product offers a connector for a popular document store. Enabling it opens a consent screen requesting read access to files. Once approved, the assistant can search and read documents, because the connector already knows how to list folders, handle paging, and refresh the token. Nobody wrote code, and what the assistant can reach was fixed by the scopes shown on that single screen. ## Related terms - [Integration](/en/glossary/integration) - [OAuth](/en/glossary/oauth) - [MCP Server](/en/glossary/mcp-server) - [API](/en/glossary/api) - [Webhook](/en/glossary/webhook) [Back to the AI Glossary](/en/glossary)