VibeKit MCP Server (vibekit-mcp) is the MCP (Model Context Protocol) server provided by VibeKit, used to deploy applications, manage hosting, and converse with AI agents in any MCP client. This package is intended for VibeKit cloud/API access and does not connect a local Claude Code instance to Telegram; for remote control of a local machine, vibekit-agent should be used.
Purpose Operate the VibeKit cloud platform through an MCP client (such as Claude Desktop, Cursor, etc.) to build applications, deploy them, manage hosting, and converse with AI agents. Each hosted application has a live URL and its own persistent server-side agent, which you can chat with.
Main capabilities After configuration, the following prompts can be used to trigger operations:
- Deploy a GitHub repository to VibeKit and create a new application
- Converse with the application's AI agent about adding a contact form
- View application logs and restart if errors occur
- Enable a database for the application, view its schema, and query the users table
- List the application's recent deployment records and roll back to the previous available version
- Run QA tests on the deployed application
- View the VibeKit account balance and list applications
- Create a weekly schedule to improve application performance
Usage
- Remote use (no installation required): VibeKit also runs as a hosted remote server, so clients that support remote MCP URLs (such as the claude.ai web connector, ChatGPT, etc.) do not require installation. Paste the URL into the client's "Custom Connector"/"Remote MCP Server" field. Clients that support setting request headers can also send Authorization: Bearer vk_your_api_key_here and use the bare address https://mcp.vibekit.bot/mcp。
- Local stdio installation (for example, Claude Desktop):
- Obtain a VibeKit API key (see below).
- Add it to the MCP client configuration (for example, Claude Desktop, with the macOS path being ~/Library/Application Support/Claude/claude_desktop_config.json).
- Restart the MCP client.
Obtaining an API key Keys start with vk_, and there are two ways to obtain one:
- Instant, no registration required: a single request returns the key. curl -X POST https://vibekit.bot/api/v1/auth/register，响应示例：{"apiKey": "vk_...", "plan": "free", "credits": 0}. Save the returned key and reuse it. Each call to /auth/register creates a brand-new account, so call it only once, not on every request.
- From an account: if you are already using VibeKit, you can copy or regenerate the key in the web dashboard or in the iOS app settings.
Environment Variables The material lists an "Environment Variables" section, but does not provide specific variable names or value descriptions, so none are provided for now.
Related packages
- vibekit-cli: terminal client for VibeKit cloud workflows
- vibekit-agent: Telegram bridge for local Claude Code on this machine
Data and privacy Every tool call is forwarded to the VibeKit API over HTTPS (https://vibekit.bot/api/v1），并使用所配置的 Authentication using API keys. Tool inputs and outputs are handled by VibeKit to run applications and agents, and are subject to its privacy policy. The server itself retains no state: it does not write to disk, and this package does not collect analytics or telemetry data; API keys are read from configuration (VIBEKIT_API_KEY / the extended key field) and are only sent to vibekit.bot. Data retention, third-party sharing, and deletion are governed by the aforementioned policy. If you have questions, you can contact support@vibekit.bot.
Use cases Suitable for users who need to directly manage VibeKit-hosted applications, deployments and rollbacks, view logs, enable databases, run QA tests, converse with each application's server-side AI agent, and view account balances and create scheduled tasks within the MCP client.
Limitations This package is only for VibeKit cloud/API access, not for connecting a local Claude Code instance to Telegram; for remote control of a local machine, use vibekit-agent. The materials do not provide specific fees, pricing, platform support scope, results, or detailed usage steps; none are provided for now.
