> ## Documentation Index
> Fetch the complete documentation index at: https://docs.kipmox.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Quickstart

> Sign in, add the kipmox MCP server to your agent, and run your first analyze → fix → verify.

Get kipmox Agent running in your AI coding agent in three steps.

## Prerequisites

* **Node.js 18+** — the server runs via `npx`, nothing to install globally.
* An **MCP-capable agent** — Claude Code, Cursor, Codex, or any client that supports MCP servers.
* A **paid kipmox plan** with Agent access — the kipmox Agent tier, or Pro/Enterprise (which include it). See [Plans & Limits](/help/plans-and-limits).
* *Optional, for your custom ruleset:* the **Salesforce CLI** with the `code-analyzer` plugin and **Java 11+**. kipmox's native rules work without them — this is only needed to re-run your own Code Analyzer / PMD ruleset. See [Custom rules](/agent/tools#custom-rules).

## 1. Sign in

The easiest way is the `login` command — it uses a magic link, so you can click it on any device (handy on VDI or a remote box):

```bash theme={null}
npx @kipmox/mcp login
```

<Steps>
  <Step title="Enter your email">
    Run the command and enter the email for your kipmox account (or pass it inline: `npx @kipmox/mcp login you@company.com`).
  </Step>

  <Step title="Click the link">
    kipmox emails you a sign-in link. Open it on **any device** — the terminal polls until you click it (up to 5 minutes).
  </Step>

  <Step title="Done">
    The token is stored at `~/.kipmox/credentials.json`, and the server picks it up automatically — **no token needed in your MCP config**. Sign out anytime with `npx @kipmox/mcp logout`.
  </Step>
</Steps>

<Note>
  **Prefer an explicit token?** Set `KIPMOX_TOKEN` in your MCP server config instead (see below). An env token always wins over the stored login — useful for CI or shared machines.
</Note>

## 2. Add the MCP server

<Tabs>
  <Tab title="Claude Code">
    If you signed in with `login`, add the server with no token:

    ```bash theme={null}
    claude mcp add kipmox -- npx -y @kipmox/mcp
    ```

    Or pass a token explicitly:

    ```bash theme={null}
    claude mcp add kipmox --env KIPMOX_TOKEN=<your-token> -- npx -y @kipmox/mcp
    ```
  </Tab>

  <Tab title="Project .mcp.json">
    Add kipmox to your project's `.mcp.json` (checked in, or per-machine):

    ```json theme={null}
    {
      "mcpServers": {
        "kipmox": {
          "command": "npx",
          "args": ["-y", "@kipmox/mcp"]
        }
      }
    }
    ```

    If you're using an explicit token instead of `login`, add it under `env`:

    ```json theme={null}
    {
      "mcpServers": {
        "kipmox": {
          "command": "npx",
          "args": ["-y", "@kipmox/mcp"],
          "env": { "KIPMOX_TOKEN": "<your-token>" }
        }
      }
    }
    ```
  </Tab>

  <Tab title="Cursor / other MCP clients">
    Any MCP client that runs stdio servers works. Point it at the same command:

    ```json theme={null}
    {
      "mcpServers": {
        "kipmox": {
          "command": "npx",
          "args": ["-y", "@kipmox/mcp"],
          "env": { "KIPMOX_TOKEN": "<your-token>" }
        }
      }
    }
    ```

    In Cursor, add this under **Settings → MCP → Add new server**. Omit `env` if you used `npx @kipmox/mcp login`.
  </Tab>
</Tabs>

<Note>
  Don't commit a real `KIPMOX_TOKEN` to a shared repo. Use `npx @kipmox/mcp login` (per-developer, stored outside the repo), or inject the token from your secret store in CI.
</Note>

## 3. Run it

Restart your agent so it picks up the new server, then just ask — kipmox's tools are called automatically:

```text theme={null}
Review AccountService.cls with kipmox, fix the issues, then verify with kipmox.
```

Your agent will:

<Steps>
  <Step title="Call analyze_code">
    kipmox returns the findings — rule, line, and a remediation hint for each.
  </Step>

  <Step title="Write the fix">
    Your agent uses the findings to write the change.
  </Step>

  <Step title="Call verify_fix">
    kipmox re-checks the fix against the original and returns a verdict — **checks-passed**, **review-recommended**, **validation-failed**, or **unable-to-verify** — *before* the change is applied.
  </Step>
</Steps>

<Info>
  kipmox verifies the fix **before** it's applied. If the verdict isn't clean, a good prompt tells the agent to rewrite and re-verify until it passes — see [`verify_fix`](/agent/tools#verify-fix).
</Info>

## Configuration

All configuration is via environment variables in your MCP server's `env` block.

| Env var | Required | Description |
| - | - | - |
| `KIPMOX_TOKEN` | Only if not using `login` | Your kipmox account token. Wins over a stored `login` token when both are present. |
| `KIPMOX_BACKEND_URL` | No | Override the backend URL (default `https://api.kipmox.com`). For Enterprise/self-hosted backends. |
| `KIPMOX_SF_CLI` | No | Path/name of the Salesforce CLI binary (default `sf`), for custom-ruleset verification. |
| `KIPMOX_ANALYZER_CONFIG` | No | Explicit path to your `code-analyzer.yml`, if auto-detection doesn't find it. |
| `KIPMOX_RULE_SELECTOR` | No | Override the Code Analyzer rule selector used for your ruleset. |

<CardGroup cols={2}>
  <Card title="Tools" icon="screwdriver-wrench" href="/agent/tools">
    What analyze\_code and verify\_fix take, and the verdicts
  </Card>

  <Card title="Troubleshooting" icon="life-ring" href="/agent/troubleshooting">
    Server not showing up, sign-in, and custom rules
  </Card>
</CardGroup>