--- title: Quickstart with Cursor description: Connect Cursor to Kaviris over MCP and deploy from the agent chat. --- # Quickstart with Cursor Cursor's agent talks to Kaviris through the Kaviris MCP server, the same way Claude Code does. ## 1. Add the server Either open **Cursor Settings → MCP → Add new MCP server**, choose type **HTTP** and enter the URL: ``` https://api.beta.kaviris.com/mcp ``` or add it to `~/.cursor/mcp.json`, which works in all your projects (a `.cursor/mcp.json` in a project works in that project only): ```json { "mcpServers": { "kaviris": { "url": "https://api.beta.kaviris.com/mcp" } } } ``` You don't paste a token. An agent chat that was already open may not see the new server: start a new chat, or reload the window. ## 2. Sign in You start this, not the agent: an agent cannot sign in for you. In **Cursor Settings → MCP**, press the login button next to **kaviris**. Your browser opens. Create your account with your email address (confirm it with the code we email you) and a passkey, then allow Cursor to see your apps, create and deploy them, and set their secrets. It can never read secret values. Deleting apps and rolling back are not given to agents automatically. You approve on a page in the browser that lists those permissions, with **Yes** and **No**. It may say **Unverified app**, because anyone can register an app under any name: choose **Yes** only if you pressed the login button yourself just now. ### Signing in from your phone or another device The agent and the browser do not have to be on the same machine: you can approve on your phone while the agent runs on your laptop, or on your laptop while it runs on a server over SSH. The browser just cannot hand the sign-in back to the agent by itself. So after you choose **Yes**, `auth.beta.kaviris.com` shows **Approved — finish in your agent**: - **Continue on this computer**: when the browser and the agent are on the same machine. - **Agent on another device?**: copy the message and paste it into the chat with your agent. It opens the address in it on its own machine with a command such as `curl`, and that finishes the sign-in. An agent that cannot run commands cannot do this step: open the address in a browser on the machine where Cursor runs, or run `curl ''` in a terminal there. The address works once, only for the agent that started the sign-in, and for about 5 minutes. Paste it only into your own agent, one you started yourself: never send it to anyone. ## 3. Ask the agent to deploy In the agent chat: > Deploy this app to Kaviris. The agent reserves a name, uploads your committed source (commit first; `.env` files are never sent), starts the deploy and follows it. On the first deploy it shows you the draft [contract](../concepts/contract.md), a short list of what the app may do, and approves it. Then it gives you the address, `https://.apps.beta.kaviris.com`. If the deploy is refused, the agent gets what was found, why, and the fix. It can usually fix the problem and deploy again by itself. Every reason is explained in the [refusal list](../refusals/index.md). Libraries with known security holes never stop a deploy; the agent offers you Kaviris's ready-made [upgrade patch](../refusals/vuln_fixable.md) instead. > **One thing to check in your code:** Kaviris tells your app which port to > listen on in the `PORT` environment variable. Read it instead of writing > a port number into the code: `app.listen(process.env.PORT || 3000)` in > Node, `int(os.environ.get("PORT", "8000"))` in Python, and listen on > `0.0.0.0`. Most frameworks already do. If yours doesn't, the deploy > stops with [port_mismatch](../refusals/port_mismatch.md), and your agent > can fix it. ## 4. Secrets > Set STRIPE_SECRET_KEY from my .env. The agent gets a one-time link from Kaviris and runs a command in Cursor's terminal that sends the value from `.env` straight to Kaviris, so the key never goes through the model. Cursor may ask you to allow the command. See [Secrets](../concepts/secrets.md). ## 5. Day to day Ask in plain words: "is my app up?", "show the build log", "let it call api.openai.com", "roll back". The agent picks the right [tool](../reference/mcp-tools.md). Rolling back and deleting always need your OK. If Cursor's agent cannot run terminal commands in your setup, use the [`kaviris` command line](cli.md) yourself for deploys and secrets.