tawk.to MCP server for your AI

Yes. tawk.to's REST API is still an invite-only beta, so you request access first and approval usually lands in about 24 to 48 hours. Once you're approved, create your key in the tawk.to dashboard and paste it into the REST API key field. Nothing else in the setup changes.

The typed tools cover four areas: the authenticated agent who owns the key, the properties (websites and workspaces) that key can access, the chats on a property, and the tickets and conversations on a property. Everything is scoped to the single account the key belongs to, so it cannot see another organization's properties. Beyond those, one universal tawk_call tool can reach any other tawk.to REST method for that same account.

The four typed tools are all reads. Write work happens through the universal tawk_call tool, which covers the rest of the REST surface: creating properties, managing members and webhooks, widgets, Knowledge Base articles and chat statistics. Those writes run as soon as the AI calls them; for reporting and nothing else, keep teammates on read-only.

This is the thing that trips people up. In tawk.to, chats and tickets belong to a property rather than to the account as a whole, so both list_chats and list_tickets are scoped per property. In practice the agent runs list_properties first and then queries each one, which is why an account-wide question takes a couple of calls instead of one.

Four operations across four API modules load up front: Account, Chats, Properties and Tickets. That keeps the tool list short enough for a client to reason over. Anything else stays searchable, and the universal call tool is there when you need a method outside the typed four.

There are five triggers PopMCP can watch by change detection: chat started, chat ended, ticket created, ticket closed, and property added. All five work by spotting the change, which PopMCP turns into an event your agent can act on, so a new visitor chat or a resolved ticket becomes something it can pick up. Property added is the useful one if new sites get onboarded regularly and you want the agent's view to keep up.

No. PopMCP hosts it and gives you a Streamable HTTP MCP URL. There is nothing to deploy, no callback URL to register, and no infrastructure to keep running or patch.

Any client that supports a remote Streamable HTTP MCP server: Claude Desktop, Claude Web, Claude Code, ChatGPT custom connectors, Codex and the Codex CLI, OpenCode and Z.ai's ZCode, as well as editors such as Cursor, Windsurf and Zed. Your client signs in to PopMCP with OAuth; the provider credential never leaves the server.

Connect tawk.to in PopMCP first, then run claude mcp add --transport http popmcp https://app.popmcp.com/mcp in a terminal. Inside Claude Code, run /mcp and sign in to PopMCP in the browser window it opens, ticking tawk.to on the consent screen. The tawk.to tools are then available in that session, and no tawk.to credential is stored on your machine.

Yes. The tawk.to catalog is small enough that all 4 tool definitions load into every conversation, so your AI never has to search for one before it can call it.

A tool is something your AI calls, and there are 4 here. A trigger is something PopMCP watches for you: 5 tawk.to events that can start work on the PopMCP side. Triggers never enter your AI client's context, and none of them ask you to paste a callback URL into tawk.to.