DeployVera · Direct AI integration

Deploy with AI: connect Claude, Cursor and VS Code directly

DeployVera runs a native, hosted MCP server. Connect Claude, Claude Code, Cursor, VS Code, Codex or any MCP client; the agent lists your projects, deploys them, follows the build and explains why a build failed. Setup is one command or one click; there is no separate bridge or server for you to run. This page gathers the connection points for each client; the details live in the guides it links to.

Last reviewed:

What the agent can do

  • List your projects and databases, and read a project’s details.
  • Deploy, follow the build, and cancel a build.
  • Read build and runtime logs, and explain in plain language why a build failed.
  • Read database logs (they carry no credentials).
  • List, add and update environment variable names; values are never shown back to the agent.
  • Create a project from a connected Git repository in a plan you already have that has a free application slot (always with your approval in the browser).
  • Give you the DNS records a custom domain needs (no approval), then attach or remove the domain once DNS is verified (always with your approval) and check the domain status.
  • Roll back to the previous version and delete environment variables: these also always need your approval in the browser.

What the agent cannot do

  • Buy or change plans, make payments, or change billing or account settings (these stay in the panel).
  • Use administrator functions.
  • Read the values of environment variables.
  • The agent always acts with the authority of the signed-in account or token, never more.

Which client, and how to connect

The server address is the same for all of them: https://mcp.deployvera.com/mcp (Streamable HTTP). The rows below are the setup paths DeployVera documents.

ClientHow to connectLink or command
Claude (web and desktop)The DeployVera connector in the Claude directoryhttps://claude.ai/directory/connectors/deployvera
Claude CodeOne commandclaude mcp add --transport http deployvera https://mcp.deployvera.com/mcp
CursorOne-click install linkhttps://cursor.com/install-mcp?name=deployvera&config=eyJ1cmwiOiJodHRwczovL21jcC5kZXBsb3l2ZXJhLmNvbS9tY3AifQ%3D%3D
VS CodeAdd the server to the MCP configuration (mcp.json)Step 3 below
CodexServer and personal token in config.tomlStep 4 below
Any other MCP clientAdd the server address as Streamable HTTPhttps://mcp.deployvera.com/mcp

Connect in four steps

On first connection you sign in in the browser and choose which permissions (read, deploy, environment variables) the agent gets and for how long. For browserless setups such as CI there is a personal token (Panel > Settings > AI agent access).

  1. 1.Claude: add it from the directory

    Open the DeployVera connector in the Claude directory and connect it: https://claude.ai/directory/connectors/deployvera

  2. 2.Claude Code: one command

    Run this in your terminal.

    claude mcp add --transport http deployvera https://mcp.deployvera.com/mcp
  3. 3.VS Code: mcp.json

    Add the server to the VS Code MCP configuration (mcp.json).

    {
      "servers": {
        "deployvera": {
          "type": "http",
          "url": "https://mcp.deployvera.com/mcp"
        }
      }
    }
  4. 4.Codex: config.toml and a token

    Create a personal token in the panel, put it in the DEPLOYVERA_TOKEN environment variable and add this to the configuration. For Cursor, the one-click install link above is enough.

    [mcp_servers.deployvera]
    url = "https://mcp.deployvera.com/mcp"
    bearer_token_env_var = "DEPLOYVERA_TOKEN"

Safety model and details

  • Sign-in happens in the browser and you choose which of read, deploy and environment variables to grant. Rollbacks, deleting environment variables, creating a project and attaching or removing a domain always need approval; you can also require approval for every deploy.
  • Approval is given in the browser with your account password and a two-factor code; an approval link is single-use and expires after 10 minutes.
  • Tools that need approval: `rollback`, `delete_env_var`, `create_project`, `set_domain` and `remove_domain`. `start_domain_verification` only returns the DNS records and needs no approval; `check_domain`, `list_services` and `list_repositories` are read-only.
  • For step-by-step setup, an example session and a FAQ see the Claude Code guide; for technical detail see the MCP article. This page does not repeat them.

Source

Frequently asked questions about the AI integration

How do I deploy my app with AI?

Connect your client (Claude, Claude Code, Cursor, VS Code, Codex or another MCP client) to the DeployVera MCP server at https://mcp.deployvera.com/mcp. After you push your code, tell the agent to deploy the project; it starts the deployment, follows the build and reports the result.

Which clients work?

DeployVera documents Claude (web, desktop, Claude Code), Cursor, VS Code, Codex and any MCP client. No guarantee is given for a client not listed here; if it supports Streamable HTTP you can add the server address.

Can the agent delete, roll back or create a project without asking me?

No. Rollbacks, deleting environment variables, creating a project and attaching or removing a domain always need your approval in the browser with your password and a two-factor code. You can also require approval for every deploy.

Can the agent see my secrets?

No. The agent can add and update environment variables but their values are never shown back to it; database logs carry no credentials either.

Can the agent create projects, connect a domain or pay?

It can create a project and attach a domain; both always need your approval in the browser. A project is created only in a plan you already have with a free application slot. The agent cannot buy or change plans, make payments, or change billing or account settings; those stay in the panel.

Does it work on the free plan?

The AI integration works with the projects in your account; the free plan includes 1 app, 0.5 vCPU, 512 MB RAM and 2 GB storage, needs no card and does not sleep when idle.

Create a free account first, then connect your AI.

No card needed; let your agent deploy the projects in your account.

Start free