> For the complete documentation index, see [llms.txt](https://docs.sonarsource.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.sonarsource.com/sonarqube-cli/integrations/claude-code.md).

# Claude Code

Install secrets-detection hooks, the SonarQube MCP Server, and Vortex analysis and Vortex context features for Claude Code with the SonarQube CLI.

`sonar integrate claude` configures the SonarQube CLI to work alongside [Claude Code](https://claude.com/product/claude-code). In an interactive terminal, the command shows these recommended components, then asks you to install them or customize the selection (see [Install confirmation](/sonarqube-cli/integrations/integrations.md#install-confirmation)):

* **Secrets-detection hooks**: a `UserPromptSubmit` hook that scans the prompt you send to Claude, and a `PreToolUse` hook that runs before Claude reads or writes files. Together they block any operation that would expose a secret.
* **The** [**SonarQube MCP Server**](https://docs.sonarsource.com/sonarqube-mcp-server/), which lets Claude fetch projects, issues, and rules directly.
* Vortex (when your connection has a Vortex subscription: SonarQube Cloud Team (annual) or Enterprise plan, or SonarQube Server 2026.5 or newer Enterprise and Data Center editions): installs Vortex analysis and Vortex context together. It adds:
  * A shared [Vortex analysis](/agent-centric-development-cycle/inside-your-agent-the-agentic-loop/sonar-vortex-analysis.md) `PostToolUse` hook that runs fast `STANDARD` analysis after each edit or write and forwards Bash, PowerShell, Monitor, and Read tool results to Vortex context when those hooks are available
  * A `PostToolUseFailure` hook that forwards failed Bash, PowerShell, Monitor, and Read tool calls to Vortex context when those hooks are available
  * End-of-turn Vortex analysis instructions in `CLAUDE.md` that tell Claude to run `DEEP` analysis on every file it modified in one command (multi `--file` or git change set)
  * `SessionStart` (startup and clear) and `SubagentStart` hooks that inject [Vortex context](/agent-centric-development-cycle/inside-your-agent-the-agentic-loop/sonar-vortex-context.md) when a session starts, letting Claude use project guidelines, architecture, semantic navigation, and dependency context

Pass `--non-interactive` to accept every offered component without prompts (see [Non-interactive install](#non-interactive-install)).

## Prerequisites

* [The SonarQube CLI is installed and authenticated](/sonarqube-cli/quickstart-guide.md).
* [Claude Code](https://docs.claude.com/en/docs/claude-code/setup) is installed.

## Install

Install once for this machine:

```bash
sonar integrate claude
```

The command writes hooks and config to `~/.claude/` and applies to every Claude Code session on this machine. See [Project versus global scope](/sonarqube-cli/integrations/integrations.md#project-versus-global-scope).

> **Note:** If you integrated Claude Code before Vortex context hooks were added, update to SonarQube CLI 1.8.0 or newer, then run `sonar integrate claude` again to install them. Re-running also replaces a leftover Vortex context skill file (`SKILL.md`) from an earlier CLI version with the session-start hooks.

### What the command does

The integrator runs in two phases:

1. Validation. It checks that your saved token can reach SonarQube. On SonarQube Cloud, your connection must include an organization. An invalid token stops setup and tells you to run `sonar auth login`. An unreachable server stops setup and asks you to check the server URL and your network connection.
2. Installation. The CLI shows a [What will be installed](/sonarqube-cli/integrations/integrations.md#install-confirmation) summary of the recommended components (secrets hooks, MCP server, and Vortex when eligible). In an interactive terminal, choose **Install recommended** to write that set, or **Customize** to accept or decline each component and then press Enter. Components that are not available are skipped with an explanation. In non-interactive mode, every eligible component is installed without prompts. The CLI then writes the accepted components into `~/.claude/`.

   Common skip reasons include:

   * Your connection does not have a Vortex subscription (only available with SonarQube Cloud Team (annual) or Enterprise plan, or SonarQube Server 2026.5 or newer Enterprise and Data Center editions).

### Options

| Option              | Description                                                                                                                        |
| ------------------- | ---------------------------------------------------------------------------------------------------------------------------------- |
| `--non-interactive` | Non-interactive mode (no prompts).                                                                                                 |
| `--global`, `-g`    | Deprecated since 1.9.0. The install is already global; using the flag prints a warning and does not change the installation scope. |

### Where files are installed

| Location        | Hook and config                                                                             |
| --------------- | ------------------------------------------------------------------------------------------- |
| User home       | `~/.claude/` (hooks, settings, MCP config, and `CLAUDE.md` Vortex analysis instructions)    |
| Shared binaries | `~/.sonar/sonarqube-cli/bin/` (auxiliary binaries for secrets detection and Vortex context) |

State for installed integrations is recorded in `~/.sonar/sonarqube-cli/state.json`. See [State and storage](/sonarqube-cli/administration/state-and-storage.md).

## Restart Claude Code

Claude reads its hook and MCP configuration at startup. Restart Claude Code for the integration to take effect.

## Vortex analysis and Vortex context

When you run `sonar integrate claude` against a SonarQube Cloud project, Vortex installs these Claude Code touchpoints:

* A single `PostToolUse` hook handles both Vortex analysis and Vortex context. After each edit or write, it runs fast `STANDARD` analysis and surfaces findings inline. After Bash, PowerShell, Monitor, or Read tool calls, it forwards results to Vortex context when those hooks are available. The CLI auto-detects the current git branch for each analysis when `--branch` is not passed.
* A `PostToolUseFailure` hook forwards failed Bash, PowerShell, Monitor, and Read tool calls to Vortex context when those hooks are installed.
* Instructions in `CLAUDE.md` require a mandatory end-of-turn `DEEP` run: one command covering every file Claude modified that turn (repeat `--file` for each path, or fall back to the git change set with `--depth DEEP`). Those instructions go in `~/.claude/CLAUDE.md`. Branch context is auto-detected, and each file is analyzed against the project that owns it, so agents do not need to pass `--branch` or `--project`. In a repository with several projects, the instructions tell Claude to run that command from a directory that contains every file changed in the turn, usually the repository root. Re-run `sonar integrate claude` so an existing snippet picks up that paragraph. The post-edit hook analyzes the edited file against the project that owns it, and stays silent when no project of the active connection owns the file. If analysis reports that no project is configured, the instructions tell the agent to stay silent for the rest of the session instead of repeating a skip notice. See [Files and the project root](/sonarqube-cli/analysis/analyzing-local-changes.md#files-and-the-project-root).
* `SessionStart` (startup and clear) and `SubagentStart` hooks inject Vortex context when a session starts, letting Claude use project guidelines, architecture, semantic navigation, and dependency context. When context cannot load, Claude Code shows the warning in the UI once per session. A subagent start does not show it. See [When Vortex context does not load](/sonarqube-cli/integrations/integrations.md#when-vortex-context-does-not-load).

For overviews of these features, see [Vortex analysis](/agent-centric-development-cycle/inside-your-agent-the-agentic-loop/sonar-vortex-analysis.md) and [Vortex context](/agent-centric-development-cycle/inside-your-agent-the-agentic-loop/sonar-vortex-context.md).

## Verify it works

### Test the secrets hook

1. Create a file in your project with a fake-looking but secret-shaped value:

   ```javascript
   // secrets.js
   const API_KEY = "sqp_1aa323ae0689cd4a1abd062a2ad0a224ae8a1d13";
   ```
2. Ask Claude to read it: *"Read secrets.js."*
3. Claude Code should block the read and explain that the file contains a secret.

Once you've confirmed the hook is active, delete the test file.

### Test the MCP server

In Claude Code, ask: *"Use the SonarQube MCP Server to list my open projects."* Claude should call SonarQube and return your project list. If it does not, run `sonar auth status` to confirm the underlying token is healthy and restart Claude.

### Test Vortex analysis

In Claude Code, ask: *"Run `sonar analyze --staged` and summarize new issues."* Claude should invoke the CLI and report findings. This requires a Sonar Vortex subscription.

## Non-interactive install

For provisioning scripts, dotfiles, and onboarding automation:

```bash
sonar integrate claude --non-interactive
```

In non-interactive mode the CLI does not prompt for feature selection, install confirmation, token repair, or feature removal; it installs every component that is not explicitly skipped. When you authenticate with [environment variables](/sonarqube-cli/using-sonarqube-cli/environment-variables.md), integrate commands also run in non-interactive mode even without the flag. Run [`sonar auth status`](/sonarqube-cli/using-sonarqube-cli/commands.md#sonar-auth-status) afterward to confirm everything is configured correctly.

## Uninstall

Run `sonar integrate claude` again in an interactive terminal and choose **Customize**. For each installed feature, the CLI asks **Keep?**: answer No, confirm **Proceed with removal?**, and the CLI removes hooks, instructions, and state for that feature. Vortex is removed as one unit (Vortex analysis and context hooks, `CLAUDE.md` instructions, and session-start hooks). See [integrations/README.md](/sonarqube-cli/integrations/integrations.md#removing-installed-features) for the full flow.

For a manual cleanup or non-interactive environments:

1. Delete the hook files written under `~/.claude/`.
2. Remove the SonarQube MCP Server entry from your Claude configuration.
3. Remove the integration entry from `~/.sonar/sonarqube-cli/state.json` (under `integrations.installed`).

See [Uninstalling](/sonarqube-cli/administration/uninstall.md) for the full removal procedure.

## Related pages <a href="#related-pages" id="related-pages"></a>

* [Secrets detection](/sonarqube-cli/analysis/secrets-detection.md)
* [Analyzing local changes](/sonarqube-cli/analysis/analyzing-local-changes.md)
* [State and storage](/sonarqube-cli/administration/state-and-storage.md)


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.

Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:

```
GET https://docs.sonarsource.com/sonarqube-cli/integrations/claude-code.md?ask=<question>&goal=<user_goal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.