MCP server
The Lyvica MCP server exposes Lyvica as tools for any Model Context Protocol client. Point your agent at it and it can build and iterate on real apps — “build me a landing page for a coffee shop, then make the hero dark.”
It’s a thin wrapper over the REST API: all the logic, auth, and quota live on Lyvica’s side; the server just forwards calls using your API key.
| Tool | What it does |
|---|---|
lyvica_create_app | Create an app from a brief and queue its build |
lyvica_project_status | Get a project by id/slug — poll until READY |
lyvica_send_message | Send a follow-up instruction to iterate |
lyvica_list_projects | List your projects |
-
Create an API key at lyvica.com/dashboard/settings.
-
Get the server (it lives in the
mcp/folder of the Lyvica repo) and install its dependencies:Terminal window cd mcp && npm install -
Register it with your MCP client. For Claude Desktop (
claude_desktop_config.json):{"mcpServers": {"lyvica": {"command": "node","args": ["/absolute/path/to/lyvica/mcp/index.mjs"],"env": { "LYVICA_API_KEY": "lyv_live_…" }}}}Cline, Cursor, and other clients use the same
command,args, andenv.
Configuration
Section titled “Configuration”| Env var | Required | Default |
|---|---|---|
LYVICA_API_KEY | yes | — |
LYVICA_BASE_URL | no | https://lyvica.com |
How an agent uses it
Section titled “How an agent uses it”A typical run: the agent calls lyvica_create_app with your brief, polls
lyvica_project_status until it’s READY, then uses lyvica_send_message to
refine — each app a real GitHub repo with a live preview, on your account.