> ## Documentation Index > Fetch the complete documentation index at: https://docs.coffeechats.me/llms.txt > Use this file to discover all available pages before exploring further. # Coffee Chats documentation > Help for the Coffee Chats Chrome extension, applying to jobs, plans and Coffee Credits, and Coffee Chats MCP for AI agents like Claude and ChatGPT. Coffee Chats is an AI networking platform that helps students reach alumni and professionals with personalized outreach and turn conversations into career opportunities. These docs cover the [Chrome extension](https://docs.coffeechats.me/extension/overview.md), [applying to jobs](https://docs.coffeechats.me/extension/apply.md), [plans and Coffee Credits](https://docs.coffeechats.me/plans-and-credits.md), and Coffee Chats MCP, which the rest of this page introduces. ## Coffee Chats MCP MCP stands for Model Context Protocol. It is an open standard that lets one program ask another service to run a single named action, such as reading your contacts. Each named action is called a tool. Coffee Chats MCP lets a program such as Claude call those tools on your Coffee Chats account for you, and that program is called an agent. This page defines the words the rest of the documentation uses, gives the addresses you need, and points you at the page that answers your question. After reading it you will know what a connection is, what it can reach, and where to start. A connection carries only the permissions you granted. You can end a connection at any time. > [!NOTE] > > Across the whole tool catalogue, three tools send an email and three tools spend credits. One sends a thank-you message you already drafted, to the address stored on the contact linked to that meeting, and only after you read the exact message in your browser and approve it. One sends one outreach draft you already wrote. One queues a table of drafts for contacts you already have, after each row is checked. An agent cannot invent a recipient. One tool that spends credits starts research on your contacts, and only after you approve the batch, its cost and your balance in your browser. The other two write a resume, or a tailored resume and cover letter, and use credits as soon as the agent calls them, under their own permission. The [quickstart](https://docs.coffeechats.me/mcp/quickstart.md) has the setup steps. Six applications can connect: Claude on the web and Claude Desktop, the Claude Code command-line tool, ChatGPT, the Codex command-line tool, Cursor web agents and the Cursor desktop app. The [compatibility page](https://docs.coffeechats.me/mcp/compatibility.md) lists each one with the client identifier it uses. ## The words this documentation uses The same word means the same thing on every page. | Word | What it means | | - | - | | Connector | The Coffee Chats side of a connection. It is the program that answers an agent's tool calls at the connector address below. | | Agent | A program that does work for you by calling tools instead of using the Coffee Chats screens. | | Agent host | The application the agent runs inside, such as the Claude Code command-line tool or Claude on the web. | | Permission | One named thing you allow a connection to do, such as reading your contacts. | | Connection | One agent host's authorized access to your account, holding the permissions you granted it. | | Authorize access page | The Coffee Chats page where you grant or refuse the permissions an agent host asked for. It is headed "Authorize access". | | Review request | One change an agent has proposed and cannot make on its own. It waits until you approve it or decline it in your browser. | | Approval | Your decision in a browser about one review request. | | Credit | The unit Coffee Chats meters its paid features in. Your balance lives in the app, and the billing tools can read it. Research spends credits after you approve it, and writing a resume or cover letter spends them when the agent asks, under its own permission. | | Call | One message an agent sends to the connector over the network. This documentation never calls that a request, so that "review request" always means the thing you approve. | ## Service endpoints You use the first two addresses yourself. Your agent host reads the last two on its own when it signs you in, and you never have to open them. | Purpose | URL | | - | - | | MCP connector | `https://mcp.coffeechats.me/mcp` | | Your connections page | `https://mcp.coffeechats.me/connections` | | Protected resource metadata | `https://mcp.coffeechats.me/.well-known/oauth-protected-resource/mcp` | | Authorization server metadata | `https://mcp.coffeechats.me/oauth/mcp/.well-known/openid-configuration` | The last two addresses hold configuration that your agent host reads automatically during sign-in. They contain no account data, and you do not need to understand them to use the connector. Give the connector address to your agent host. Open the connections page yourself to see what is connected and to end a connection. ## What reading this documentation does Reading these pages installs nothing. It does not start sign-in, and it does not grant an agent anything on your behalf. Only the Authorize access page does that, and only after you sign in to Coffee Chats yourself. ## Next steps > **[Connect an agent](https://docs.coffeechats.me/mcp/quickstart.md)** > > Add the connector, sign in, choose what to share, and make a first call. > **[What an agent can do](https://docs.coffeechats.me/mcp/agents-and-tools.md)** > > What the tools reach, and the parts of Coffee Chats no tool reaches. > **[Permissions](https://docs.coffeechats.me/mcp/permissions.md)** > > How the Authorize access page works, and what each permission buys. > **[Tool reference](https://docs.coffeechats.me/mcp/tools/overview.md)** > > Every tool, generated from the connector's own catalogue. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.