Connector directory
Connector directory
Answers
Claude connectors are the integrations that let Claude reach data and tools outside its own chat window. They come in two forms: Anthropic's pre-built connectors, which wire Claude to a fixed set of services, and MCP servers, which expose any tool or data source through a standard protocol. When you ask a question, Claude calls the connector, the connector reads the source in place, and the answer comes back grounded in that live data instead of the model's memory.
Related pillar: Connect Claude to your business data
A connector is the bridge between Claude and a system it cannot see on its own. Anthropic ships a set of pre-built connectors that link Claude to specific services out of the box, so a user can grant access and start asking questions without building anything. The second kind is an MCP server, a small service that exposes a tool or data source through the Model Context Protocol, an open standard. Both solve the same problem; the pre-built connectors are turnkey for a fixed list, and MCP servers cover everything else, including your own internal systems.
The connector is the reason an answer can be about your data instead of the model's memory. When you ask Claude something, the model decides a connected tool is needed, calls the connector with a read request, and the connector queries the source in place. The result travels back into the conversation, and Claude writes the answer from what it just read. Because the read happens at question time, the answer reflects the current state of the tool, not a snapshot taken hours ago.
They overlap but are not the same thing. Claude connectors is the broader term for anything that connects Claude to an outside system, including the pre-built ones Anthropic maintains. MCP is a specific protocol that many of those connections use under the hood. A pre-built connector may be implemented with MCP or with a proprietary integration; an MCP server you run yourself is a connector you own and control. For most people the distinction matters less than the outcome: Claude can read a tool it could not see before.
A connector is only as trustworthy as the access it is given, so check three things before you hand Claude a data source. Scope, which records the connector can reach, and whether it is limited to read access. Revocation, whether you can cut the connection off from the tool itself without help. And retention, what the connector and its operator keep after a query. Read access, revocable tokens, and a clear retention answer are the three signs the integration is safe to connect.
Claude is a natively supported assistant in CorpusIQ, so the same 40+ connectors that serve ChatGPT and Perplexity also serve Claude. CorpusIQ reads the systems you approve at question time, returns the answer with a citation on each figure, and takes no actions inside the connected tools. If you already use Claude and want it answering from your live business data, this is the path.
40+ read-only connectors. ChatGPT, Claude, and Perplexity. Solo $29.95 a month. 30-day free trial.