konnektor

in English: connector

basicspreading

A connector is an interface through which an AI assistant accesses an external service – such as your emails, calendar, or files – and can work there.

In plain terms

An AI assistant only knows what it learned during training, what you type into the chat window, and what it can find on the public web. It cannot see your email inbox, calendar, or company folders. The connector opens this wall: a switchable link between the assistant and a specific service.

It is like the wall socket in an apartment. The electricity is in the wall, the device is on the table, but nothing happens until you plug it in. The socket is standardized, so you do not need to install a separate wire for every device. The AI connector is similarly a ready-made connection point: no programming is needed, just switch it on and grant permission.

The difference in practice is significant. Without connectors, the assistant only gives advice, and you must copy the texts yourself. With connectors, it retrieves last week's email, checks when you are free, or enters an appointment in your calendar. The connector is an important step towards the chat program functioning as an agent, meaning it does not just talk but also performs tasks.

How it works

You select the connector in the assistant settings (for example, Gmail, Google Drive, Slack), or the workplace administrator enables it for everyone.
You log in to the external service and approve what the assistant can access—just as when you grant an app permission on your phone.
The connector informs the assistant of the operations it offers: for example, searching emails, opening files, or creating events.
When you request something, the model decides whether an operation is needed, sends the request through the connector, and incorporates the received data into its response.
For more sensitive actions (sending emails, deletion, modification), the assistant typically asks for your confirmation first. You can revoke access at any time.

An everyday example

You ask the assistant: "When am I free next week for a one-hour meeting with Anna, and what did she write about last time?" If the calendar and email connectors are enabled, the program checks your calendar and retrieves Anna's latest email. You do not need to copy anything for it.

Who is affected

employeeWith connectors, the assistant works from your own emails, documents, and calendar, so it truly saves time rather than just giving general advice. → Work & HR

entrepreneur, executiveYou must decide which corporate systems the AI is allowed to access and who bears responsibility for this.

developerIf your product or internal system gets a connector, AI assistants and agents can use it directly. → Code

everyoneAnyone who enables a connector gives access to their personal data, so it is worth knowing exactly what they are allowing.

A common misconception

MythIf I plug in a connector, the AI immediately reads and learns all my emails and files.

In realityThe connector is an access capability. The assistant typically requests data only when needed for a task and accesses only what you have authorized within your own permissions. Whether the service provider retains the retrieved data or uses it for training is determined by the terms of that specific service. It is worth checking this.

Limits and risks

The connector sees only what you allow it to: if you grant access to your entire mailbox, confidential emails could end up in a response.
Text from an external source (such as an email from a stranger) may contain a hidden instruction for the AI. This is called prompt injection, and there is currently no complete defense against it.
Not every connector supports everything: many can only read, and if the underlying service changes or goes offline, the connector may fail.
For workplace data, do not enable a connector on your own initiative: company policies and data protection regulations (such as the GDPR) determine what data may be shared with an external AI service.

Similar, but not the same

API – An API is a raw technical interface between programs used by developers; a connector is a ready-made, click-to-activate connection built on top of this.
Model Context Protocol (MCP) – MCP is the standard describing how AI should communicate with external systems; the connector is the specific connection to a given service built on this (or another solution).
plugin – A plugin is a broader concept: it can add any new capability to a program, whereas a connector specifically links to the data and operations of an external service.

Related concepts

Broader

integration, tool use, AI assistant

Narrower

MCP-szerver, custom connector

On the edge

Model Context Protocol, API, extension, agent, prompt injection

A brief history

The term "connector" originates from the world of enterprise software, where systems-interconnecting plug-and-play components have long been called this. It spread among AI assistants after Anthropic published the open Model Context Protocol (MCP) standard in November 2024, on which increasingly more connectors are being built. At the end of 2025, the stewardship of the MCP was entrusted to an independent foundation operating under the Linux Foundation.

Where things stand

At the end of 2026, most major AI assistants offer connectors to the best-known email, calendar, file-storage and work-management services. Some providers call these by other names, such as apps or integrations. Everyday use in companies is growing rapidly, but security and permission practices are still evolving.

Status: 2026 November

In the news

Appears in 9 Agent Age stories; the curve shows weekly occurrence in the news magazine.

Epic's market dominance shapes the prospects of health care AI companiesOct 28, 2026 · Health
Epic explores AI agents for routine visits and launches Connectors to link its AI tools with third-party appsOct 20, 2026 · Health
TikTok builds an agentic shopping assistant, direct checkout and an AI ad networkOct 19, 2026 · Marketing
Harvey adds integrations and Intake product for in-house legal teamsOct 19, 2026 · Jog
Worldline, Sokin and PayNearMe launch agent-based payment toolsSep 14, 2026 · Money
Consumer AI assistants can access health records: ChatGPT and Meta's MuseSep 8, 2026 · Health
Meta AI can analyze ad campaigns, connect outside AI agents via MCP and test auto-DM replies to ad commentsAug 24, 2026 · Marketing
OSL, Cloudflare, Orla and Sokin launch stablecoin payment products for AI agentsAug 10, 2026 · Money

Also known as

konnektor · connector · AI-konnektor · MCP-konnektor · connector · connector · app connector

Sources

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.