Meta AI Connector Terms Explained: MCP, API, Developer Preview and More

A connector is a link that lets an AI assistant use another service, usually through that service's API. MCP is a common standard for building such links, and 'developer preview' means Meta's own early-access label. You can use a connector without understanding the plumbing, but a few questions to a provider will tell you whether it is ready.

By Dance Technologies · Updated 2026-11-08

The terms in one list

Each entry gives a short definition, then what it means for you. Where a term comes from Meta's own wording, we say so. Where it is a general industry meaning, we say that too.

API. A set of rules that lets one piece of software ask another for information or actions. *For you:* an API is the doorway a connector uses. If a service has no API, there is nothing for a connector to call.
MCP (Model Context Protocol). A general-purpose way for AI assistants to talk to outside tools. *For you:* it is a technical standard, not a product you buy. Ask a provider whether their connector uses it, and why that matters for your setup.
MCP server. A program that offers a set of tools to an AI assistant using MCP. *For you:* this is what a provider runs on its side. You should not have to host or manage one.
Connector. A link between an AI assistant and an app or service. Meta Help Center says you can connect Muse to apps and services from Meta or third parties, and calls these Connectors. *For you:* it is how an assistant reaches your tools, or how your service reaches the assistant.
Meta AI Connectors. In Meta's developer wording, a way to connect your service directly into Meta AI so people can use it without leaving the conversation. *For you:* this is aimed at companies that offer a service, and it is a different thing from connecting your own Instagram or ad account.
Developer preview. Meta's label for the availability of Meta AI Connectors on its developer FAQ. *For you:* expect early-stage access that may change. More below.
Muse. Described on Meta's site as a personal AI agent that can connect to apps through Connectors. *For you:* Muse is where many business connections show up, including those under Muse for Small Business.
Agent. An AI system that carries out tasks, not just answers questions. *For you:* an agent does things, a connector gives it access to things. They are not the same part of the stack.

API and MCP: what a business owner needs to know

An API is how software talks to software. When Meta's developer FAQ says a connector 'turns your API into tools Meta AI can call,' it means a company that already has an API can describe a set of actions, and Meta AI can trigger them when someone asks for that service. The same page says there is no new SDK and no new app to ship. Meta's developer FAQ is the source for that wording.

MCP stands for Model Context Protocol. It is a widely used standard that lets AI assistants discover and use tools offered by an MCP server. Other connector products use it too: one hosting product's connector, for example, is described as pointing Cursor, Claude or any MCP client at your hosting (source).

Two cautions:

The Meta pages we cite here describe Meta AI Connectors in terms of APIs and tools. They do not, in the text we reviewed, define the connector in terms of MCP. If a provider says its product is 'a Meta AI MCP connector,' ask exactly what that means and where Meta documents it.
MCP is sometimes expanded wrongly. One community post expands it as 'Meta Connect Platform' (post), which differs from the standard Model Context Protocol. Check the expansion before you rely on anyone's explanation.

What it means for you: you do not need to run an API or an MCP server yourself. You need to know who does, and what the assistant is allowed to do through it.

Developer preview: what the label tells you

Meta's developer FAQ lists Meta AI Connectors as 'Available in developer preview.' Meta's page uses the label; the explanation below is the general meaning of the phrase, not a Meta definition.

A developer preview is normally an early release aimed at developers who build with a product before it is finished. In practice that usually means:

Features, rules or access can change.
Not everyone who wants access will have it.
Documentation and support may be thinner than for a finished product.

Do not confuse it with other labels. Coverage of Meta's separate ads connectors describes an open beta that needs no developer credentials, API integrations or coding (Marketing-Interactive). A third-party write-up of Meta's Muse connector guidelines also says a connector is optional, reviewed and not guaranteed (J.A. Watte). That is a secondary account, so check Meta's current guidelines.

What it means for you: if a provider's connector depends on something in developer preview, treat its availability and features as provisional. Do not build a launch plan around a date nobody has confirmed.

Connector vs agent, and Muse vs Meta AI

A connector is a link. An agent is the thing that acts. Meta describes Muse as an agent that can answer questions, complete tasks, browse the web and connect with apps and services via Connectors (Meta Muse).

Meta also has a separate product, Meta Business Agent, which answers customer questions on your behalf, learns from your website and catalog, and lets you set guardrails. For larger businesses needing programmatic access and custom integrations, Meta points to the Business Agent Platform (Meta Business Agent). That is a different thing from a Meta AI connector. We lay out the difference in Meta AI Connectors vs Meta Business Agent Platform.

Muse and Meta AI. Meta's Meta AI page says Meta AI is powered by Muse Spark, so the names are related, but Meta presents them as separate things: Meta AI as the assistant, Muse as a personal AI agent (Meta AI). Do not assume features or connectors on one apply to the other.

Two ways the word 'connector' is used. Sometimes it means linking your own accounts (Instagram, ads, Google Workspace) to an assistant. Sometimes it means a company connecting its service into Meta AI. When you read a claim, work out which one is meant.

What to ask a provider

Use these when a provider uses any of the terms above. They are a short version of the longer buyer's checklist for website connectors.

Which assistant does it work with, and how? Ask for the exact assistant and the exact connection route, not just 'works with AI.'
What status is it in? Developer preview, beta or generally available all mean different things for reliability.
What can it read, and what can it change? Read-only access and write access carry very different risks.
Who runs the MCP server or API, and what happens to my data? Ask what is stored and where.
Do I keep control of my site and domain? Ask whether you can export your content and leave.
What still needs a human? AI-assisted output usually needs review, particularly for SEO elements such as titles and descriptions.
Is it Meta's product, or a third party's? The answer should be plain and easy to find.

On disclosure: this site is published by Dance Technologies, which makes two connectors of its own. It is not independent, not a Meta product, and has not tested or ranked providers. See About.

Questions

Do I need to understand MCP to use a connector?

No. MCP is a technical standard that providers work with behind the scenes. As a business owner you mainly need to know what the connector can read and change, who runs it, and whether you can leave. The provider questions in this guide cover those points.

Is Muse the same as Meta AI?

Meta presents them as related but distinct. Meta describes Muse as a personal AI agent that connects to apps through Connectors, and says Meta AI is powered by Muse Spark. Check which one a connector or feature actually applies to.

What does developer preview mean for Meta connectors?

Meta's developer FAQ lists Meta AI Connectors as available in developer preview. In general, that signals early-stage access aimed at developers, where features and availability can change. Treat any plan that depends on it as provisional.

Is a connector the same as an API?

No. An API is the doorway a service exposes. A connector is the link that lets an AI assistant use that doorway. Meta's developer wording says Meta AI Connectors turn your API into tools Meta AI can call.

Does a connector mean the AI can build my website?

Not by itself. The word describes a link between an assistant and a service. What the link can do depends on the specific connector. Our guide on whether Meta AI can build a website covers that question directly.