Lovable MCP Server: Control Lovable from Any AI Client
LovableLovable launched its official MCP server in research preview, available at https://mcp.lovable.dev for Pro and Business plan users. External AI clients — including Claude Desktop, Claude Code, Cursor, and any MCP-compatible tool — can now connect and manage Lovable projects entirely through natural language. The server exposes project creation, agent messaging, file inspection, SQL execution, analytics queries, and deployment, effectively making Lovable an execution layer within any AI-native workflow. Tool schemas are subject to breaking changes during the preview period.
Sources & Mentions
3 external resources covering this update
Lovable as an AI-Native Execution Layer
Lovable released its official Model Context Protocol (MCP) server in research preview, enabling developers to integrate Lovable's full app-building and deployment capabilities into the AI tools they already use. The server is accessible at https://mcp.lovable.dev and available on Pro and Business plans.
The MCP server positions Lovable not just as a standalone builder, but as a programmable execution layer within larger agent workflows. An AI coding assistant like Claude Code or Cursor can now spin up a Lovable project, send instructions to the Lovable agent, inspect the resulting code, and deploy the app — without the developer ever opening a browser tab.
What the MCP Server Enables
When connected, external AI clients can perform the full range of Lovable operations:
- Create new projects from natural language descriptions
- Send messages to the Lovable agent to build features or fix bugs
- Inspect files and diffs to review what was generated
- Execute SQL queries against Lovable Cloud databases
- Query project analytics for published apps
- Enable Lovable Cloud for a project
- Deploy applications to Lovable's hosting infrastructure
A developer working in Claude Code can issue a single instruction — "build me a CRM dashboard with Supabase auth" — and have Lovable handle the entire frontend and backend build, from within their terminal.
Research Preview Caveats
Lovable is explicit that the MCP server is in active development. Tool names, parameter schemas, and response shapes are subject to change without notice, including breaking changes. The research preview label means early adopters should expect iteration. Lovable invites feedback to shape the final stable API.
Compatible Tools
The server works with Claude Desktop, Claude Code, Cursor, Windsurf, and any other MCP-compatible AI client. Configuration is done by pointing the client to https://mcp.lovable.dev and authenticating with a Lovable account token.