> ## Documentation Index > Fetch the complete documentation index at: https://docs.voxworks.ai/llms.txt > Use this file to discover all available pages before exploring further. # MCP Connector Overview > Connect an outside AI assistant such as Claude to your Voxworks team, and what it can do for you once it is connected. ## What is the MCP connector? The Voxworks MCP connector lets an outside AI assistant work inside your Voxworks team. MCP (Model Context Protocol) is the open standard AI assistants such as Claude use to connect to other services. Once you connect one, you can ask it in plain language to do work you would otherwise do in the app: * *"Why did the last three calls on the Bookings script end early?"* * *"Change the greeting on our inbound script to mention the Easter hours."* * *"Turn this questionnaire into a survey."* (with a Word or PDF file attached) * *"Is my Monday batch stuck?"* * *"Go through the open recommendations on the Sales script with me."* [Integrations](/integrations/overview) connect Voxworks out to other services so your calls can book appointments and update CRMs. The MCP connector works the other way round: it lets an assistant connect in to Voxworks and act for you. Voxworks also has an assistant built into the app that needs no setup. See [Compared with MCP assistants](/assistant/in-app-capabilities#compared-with-mcp-assistants) for how the two differ. *** ## Who can use it Anyone who belongs to a team can connect an assistant. When you approve the connection you pick one team, and the assistant only ever works inside that team. It can never do more than your own role in that team allows, and it can only do the things you approved on the authorisation screen. See [Permissions](/mcp/permissions). *** ## What it can do | Area | What the assistant can do | Read more | | - | - | - | | **Scripts** | Read, explain and review scripts. Edit them as new versions, change a script's settings, copy a script, roll a change back, and write a brand-new script with AI, including from a script document you give it. | [Working on scripts](/mcp/scripts) | | **AI recommendations** | Go through the suggested script changes with you, apply or reject them, undo applied ones, file its own suggestions and ask for a fresh review. | [Recommendations](/mcp/recommendations) | | **Surveys** | Turn a questionnaire (Word, PDF or text) into a survey draft, check it and follow its test compile. | [Building surveys](/mcp/surveys) | | **Calls and operations** | Look into calls and their transcripts, outbound batches, automation runs, phone numbers, usage and calendar setup to explain what happened. | [Calls and operations](/mcp/calls-and-operations) | | **Help** | Look things up in these docs and in the Voxworks playbook before it answers, and follow built-in guides for common jobs. | [Built-in guidance](/mcp/built-in-guidance) | For the full list of tools, see the [Tools reference](/mcp/tools-reference). *** ## How your work stays safe * **Every change is a new version.** Saved changes never overwrite what was there, so you can always go back. See [Versions & Publishing](/scripts/versions-and-publishing). * **Changes are checked first.** A script change has to pass the same checks the script editor runs before it can be saved. The assistant previews a change before it saves it. * **Live calls only change if you allowed it.** Without the **Publish script changes to live calls** permission, the assistant's script changes are saved as drafts for you to publish in the app. New surveys are always drafts until someone publishes them in the survey editor. * **You can cut it off at any time.** Revoking an assistant under **Settings > MCP Clients** stops it within seconds. See [Managing connected apps](/mcp/managing-connections). *** ## Next steps Add Voxworks to your AI assistant and approve the connection. What each permission on the authorisation screen allows. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.