Learn AI /AI Agents and Automation /Connector
Connector
- Starter
- AI Agents and Automation
- Programme & data teams
Also called: integration, plugin.
Back to the milestone you came from
In five seconds
What it is
A connector links an AI system or app to another digital service so it can send requests, receive results, and use them in a task.
In plain words
A connector lets one digital tool reach another one so information or actions can pass between them.
An everyday example
A traveller arrives at a hotel in Nairobi and asks the receptionist for three things: a taxi to the airport, tomorrow's weather, and a dinner booking. The receptionist does not drive the taxi, measure the weather, or cook the meal. Instead, they contact the taxi company, check a weather service, and call a restaurant, then bring the results back to the traveller in one place. If one service is closed, the receptionist reports that and may ask the traveller what to do next.
A human receptionist can improvise, use judgement, and notice social context in ways software often cannot. Real connectors usually follow fixed rules, data formats, and permissions. Also, one connector does not normally handle every kind of service by personal judgement; each connection is built for a specific system or standard.
Remember thisA connector gives an AI system a defined route to another service. What it can retrieve or do should still be limited by authentication, permissions and the user's instructions.
This term is contested
The general idea is widely shared, but product teams use "connector", "integration", and "plugin" differently, so the boundary between them is not fully consistent across platforms.
Understand it
Understand it
Why it matters
For non-specialists, connectors matter because they change what an AI system can actually do. A tool without connectors may only generate text. A tool with connectors may retrieve current information, update records, create drafts from your files, or trigger actions in other systems. That can save time, but it also raises practical questions: Which systems are connected? What data can move? Who approved the connection? What happens if the outside service is wrong, unavailable, or too broadly authorised? Understanding connectors helps you ask better oversight, privacy, and reliability questions before relying on automated outputs or actions.
How a connector moves a request to another service and back
After viewing this, a reader should be able to explain that a connector passes a request from an AI system to an external service, translates formats if needed, and returns the result.
01
1. Request starts
A user or workflow asks for something, such as "show tomorrow's meetings" or "draft an email using this file".
Draw a box labelled 'User or workflow request' with a speech bubble icon and a short sample request under the label.
02
2. AI system chooses a tool path
The assistant or workflow identifies that it needs information or an action from another system.
Draw a second box labelled 'AI system' with a branch note reading 'needs outside service' and an arrow from step 1.
03
3. Connector translates and sends
The connector packages the request in the format the target service expects, using allowed credentials and settings.
Draw a third box labelled 'Connector' with small tags 'format', 'permissions', and 'send'. Add arrows both ways between this box and the next.
04
4. Service responds
The external service returns data or confirms an action, and the result comes back through the connector to the AI system and user.
Draw a fourth box labelled 'External service' with two possible outputs underneath: 'data returned' and 'action confirmed'. Add a final arrow back toward the AI system and a small result badge near step 1.
Show a left-to-right flow with four boxes and arrows. Read it from the user's request through the AI system, into the connector, then to the external service, and back with a result. Use labels to show that the connector handles the handoff and translation, while the external service provides data or performs the action.
Hovering or tapping each box highlights what that component does and dims the others; arrows brighten to show the request path and the return path.
Describe this without the diagram
The diagram is a four-step flow read from left to right. Step 1 is the user or workflow request. Step 2 is the AI system, which recognises that it needs an outside service. Step 3 is the connector, which translates the request into the right format, applies allowed permissions, and sends it onward. Step 4 is the external service, which returns data or confirms an action. The return then travels back through the connector to the AI system and finally to the user. The main lesson is that the connector is the link and translator, not the original source of the data.
With reduced motion: A static four-box flow replaces animation. Solid arrows show the path from request to AI system to connector to external service, and dotted arrows show the result returning along the same route.
A connector is a bridge between one system and another. In AI tools, it lets a chatbot, assistant, or automated workflow reach an outside service such as email, a calendar, a database, or a document store.
What goes in? Usually a request from the AI system or the user, plus any needed settings such as which account to use or what action is allowed. What happens next? The connector turns that request into the format the other service expects, sends it, and brings the response back.
What comes out? The result might be data, such as a list of meetings, or an action confirmation, such as "draft saved" or "message sent". The connector itself does not usually generate the answer text. It mainly handles the link and the data exchange.
People often confuse a connector with an API. An API is the set of rules a service offers for software to interact with it. A connector is the specific piece that uses those rules in a product or workflow. People also confuse it with a plugin. In practice, "plugin" and "connector" are often used loosely, but a plugin may also add interface features or logic, not just a connection.
How “A hotel receptionist calling nearby services for a traveller” maps onto the real thing
| In the story | In the AI system |
|---|---|
| the traveller's requests | the user's prompt or the workflow's task |
| the receptionist | the connector layer that passes requests and returns results |
| taxi company, weather service, and restaurant | external services such as maps, weather, booking, or email systems |
| the receptionist's update back to the traveller | the returned data or action status shown to the user or passed to the next step |
A worked example
An evaluation officer asks an approved drafting assistant to find the latest cleared evaluation summary held by the technical division and prepare source-linked points for a delegation briefing.
Starting from: "Find the latest cleared evaluation summary for this workstream and draft three briefing points using only that document."
- The assistant determines that it needs to search the division's document store and calls its connector.
- The connector authenticates the officer's account and maps the request to search fields for workstream, document type and clearance status.
- The document store returns the matching record, including its version, clearance status and file link. The connector normalises these fields for the assistant.
- The assistant drafts three points from the retrieved text and attaches a source link to each. It does not retrieve records the officer is not authorised to access.
- The officer checks the source, version and wording before placing the points in the briefing note for clearance.
Result: The drafting workspace shows three source-linked briefing points and identifies the cleared record used. If no cleared record is found, it reports that result rather than substituting an uncleared draft.
In a United Nations workplace
In a specialised agency's technical division, an evaluation officer is assembling a briefing note for a delegation while comments from capitals are still pending. An approved drafting tool uses a connector to query the division's document store for the latest cleared evaluation summary, return relevant passages and file metadata, and place source links in the drafting workspace.
The connector handles authentication, translates the officer's request into the document store's search parameters, and returns the results in a form the drafting tool can use. The officer does not need to search several folders or copy each passage manually.
Before you try thisThe connector must preserve the document store's access controls. Do not connect an unapproved AI service or expose confidential interview material, personal data or draft findings. The officer should confirm that the retrieved version is cleared and current before using it in the briefing note.
Where you may hear it
- During a vendor demo to an ICT governance board
- In a procurement unit's technical evaluation meeting
- While a programme officer maps a workflow with an information management focal point
- At a security review before an automation is linked to a document store
Avoid confusion
Avoid confusion
Often confused with
API
An application programming interface defines rules that software uses to request data or actions from another system.
Model Context Protocol
A standard way for an AI application to connect with external tools and data sources through a shared, predictable interface.
Function calling
A method that lets a model return a structured request for an external tool or system instead of only producing free-form text.
Computer use
A system capability that lets software observe what appears on a screen and operate apps by clicking, typing, scrolling and selecting items.
Common misconceptions
Common misconception: A connector is the same thing as the AI model.
Correction: The model generates or transforms content, while the connector links that system to another service. If a calendar entry appears in an assistant, the connector likely fetched it; the model did not create the event record from nowhere.
Common misconception: If a tool has a connector, it can safely do anything in that external system.
Correction: A connector only works within the permissions, fields, and actions it was configured to use. Some connectors are read-only, some can write, and some expose only a small part of a service. Safety still depends on access control, review, logging, and human oversight.
Common misconception: Connector, plugin, and API always mean exactly the same thing.
Correction: They are related but not identical. An API is an interface a service exposes for software use. A connector is a built link that uses such interfaces. A plugin may include a connector, but it can also add user-interface elements or extra logic inside an application.
Quick check
Which description best matches a connector in an AI workflow?
A connector is the practical link between systems. It uses rules such as an API, but it is not the API itself, and it does not guarantee that returned information is accurate or appropriate.
Go deeper
Go deeper
The simple version
A connector lets one digital tool reach another one so information or actions can pass between them.
The technical version
A connector is an integration component that mediates between a host system and an external service. It typically handles authentication, request formatting, parameter mapping, error handling, and response normalisation so the host can call functions or retrieve data without implementing each service interaction from scratch. In AI systems, connectors often sit between the model-driven layer and tools such as email, calendars, search, databases, document stores, or business applications. This is related to tool use and function calling, but not identical: function calling is a method for producing structured tool requests, while the connector is the implementation that actually reaches the target service and returns structured results.
Reviewed external resources
Primary sourceModel Context Protocol (opens in a new tab)Model Context Protocol
This official MCP resource explains one emerging standard for connecting models to tools and data, and it suits readers who want a more technical, standards-oriented view.
Deeperlink checked 2026-07-31
EncyclopaediaWikipedia search: Connector (opens in a new tab)Wikipedia
This Wikipedia link is a search results page, so it helps readers see the many non-AI meanings of "connector" and suits beginners who need context before narrowing the term.
search results, not one pagelink checked 2026-07-31
ArticleBuilding effective agents (opens in a new tab)Anthropic
This article gives a broader explanation of agents and tools, which helps readers place connectors inside agent workflows and suits readers who want a conceptual overview.
Intermediatelink checked 2026-07-31
VideoVideo explanations: Connector (opens in a new tab)YouTube search
This YouTube link is a search results page, so it helps visual learners compare different video explanations of connectors and suits beginners who prefer demonstrations.
search results, not one pagelink checked 2026-07-31
DiscussionCommunity discussion: Connector (opens in a new tab)Reddit search
This Reddit link is a search results page, so it shows how practitioners discuss connectors in real use and suits readers who want practical questions, examples, and common points of confusion.
search results, not one pagelink checked 2026-07-31
Where this came from
Drafted with a language model through an offline pipeline, checked against this site's schema, and every external link above fetched to see whether it answers. Treat it as a starting point and check anything that matters against the primary sources. How this site is built and what that means.
The one thing to remember
A connector gives an AI system a defined route to another service. What it can retrieve or do should still be limited by authentication, permissions and the user's instructions.