SonarQube MCP server: 48 tools you can narrow

A code-quality connector that treats context cost as a setting, with a read-only mode, and a header that lets a single request narrow its own tool list.

Last verified 27 August 2026 · from SonarQube's own repository README and the GitHub record for that repository

This page is one of 90 in a directory of monitoring and developer tools. Each page checks what an AI coding agent can get out of the tool through its MCP server, the connector that lets the agent query the tool directly.

What is the SonarQube MCP server?

It is SonarQube's server for the Model Context Protocol. SonarQube analyzes code quality and security, so a connected agent can check its own output and read what the platform already found.

The tools split into two kinds. Some query the platform, such as search_sonar_issues_in_projects and get_project_quality_gate_status. Others run analysis, such as analyze_file_list through a running SonarQube for IDE instance.

A third group reads code structure rather than findings. get_upstream_call_flow, get_downstream_call_flow, get_type_hierarchy and get_references are context tools, gated behind an organization entitlement.

SonarQube's 48 documented tools examine analysis, coverage, risks, issues and quality gates. Those are software-delivery signals; this server does not report production behavior, agent traces or cost.

For an agent, that means SonarQube answers whether the code meets a standard and where it does not. It can read issues, check a quality gate and trace a call flow. It knows nothing about the running system, so a question about behavior needs a telemetry connector.

What can an AI agent do with SonarQube?

Start with search_my_sonarqube_projects. It returns the project key every other tool needs, so the agent discovers the key first and composes the issue or gate query second. Its projects toolset is always enabled, and nothing in SONARQUBE_TOOLSETS turns it off.

What connecting costs before the first question

A context window is the amount of text a model can hold at once. Forty-eight tools across nineteen categories cost 3,554 to 4,446 tokens on Claude at connection. That is the largest documented surface of any code-quality connector here.

This is one of the few servers where the number is adjustable. A read-only switch and toolset filtering both narrow what registers, so the figure below is the ceiling rather than the only option.

Encodings: GPT-5.6, GPT-5.5 o200k_base; GPT-4, GPT-3.5 cl100k_base; Claude Haiku 4.5 to Opus 5 count_tokens. Counted from the 48 tool entries in the vendor's own documentation. That is the only public inventory.

The server could not be run for a live tools/list. So no input schemas are available. These figures cover each tool's name and description only. They are a floor, not the figure.

The Claude columns are measured through the API's count_tokens endpoint, with an empty schema per tool. So they cover the same content as the GPT columns. The range there is two tokenizer generations. Haiku 4.5 and Opus 4.6 sit at the low end. Opus 5 and Sonnet 5 sit at the high end. The gap between them is about a quarter.

Can SonarQube watch your own AI agents?

The word agentic appears in this connector, and it describes your codebase rather than your agents. Read from SonarQube’s MCP documentation on 24 August 2026.

SonarQube MCP server limits: where answers come back incomplete

SonarQube treats its tool list as configuration rather than a fixed fact, creating the limits below.

How to configure SonarQube MCP for agents

1. 1

   Choose the toolsets deliberately

   SONARQUBE_TOOLSETS replaces the default selection entirely, so naming one toolset switches every other off except projects.

2. 2

   Turn on read-only mode unless writing is the point

   SONARQUBE_READ_ONLY disables issue status changes and every other write, and its effect is cumulative with the toolset filter.

3. 3

   Know which mode you need

   Vortex context and analysis tools are documented as available only in stdio mode, and both need an organization entitlement.

4. 4

   Pick Cloud or Server correctly

   SonarQube Cloud needs an organization key, SonarQube Server needs a URL, and the US Cloud region needs its own URL as well.

Paste this into your agent’s instructions

The tool list depends on settings, so an absent capability can mean a disabled toolset rather than a missing feature. The block makes the agent say which.

When you answer using SonarQube MCP tools, state:
- The project key and branch the answer covers.
- Whether a capability was unavailable because its toolset is
  not enabled, rather than because the platform lacks it.
- Whether you called change_sonar_issue_status or
  change_security_hotspot_status. Both change triage state.
Never state that data does not exist. State that no issues
matched the project and filters above.
Write your answer in ASD-STE100 Simplified Technical
English. Use short sentences with one idea in each.

Do you need the MCP server at all?

Not inside a pipeline. SonarQube has a web API, and for a dashboard or a gate check in a pipeline that is the direct route.

The connector earns its place through the analysis and context tools. Analyzing a file through SonarQube for IDE, or tracing a call flow, is not something a REST call returns.

For scanning without a platform at all, a local scanner such as Semgrep answers a narrower question with no server behind it.

Connect it when the agent works inside a SonarQube project. Use the web API for gates and reporting.

SonarQube MCP server setup

Every documented configuration runs the container image, so a container runtime is a prerequisite. There is no package or standalone binary.

SonarQube Cloud needs a token and an organization key. SonarQube Server needs a token and the server URL instead.

claude mcp add sonarqube \
  --env SONARQUBE_TOKEN=$SONAR_TOKEN \
  --env SONARQUBE_ORG=$SONAR_ORG \
  -- docker run --init --pull=always -i --rm \
     -e SONARQUBE_TOKEN -e SONARQUBE_ORG \
     sonarsource/sonarqube-mcp

{"mcpServers":{"sonarqube":{"command":"docker",
  "args":["run","--init","--pull=always","-i","--rm",
    "-e","SONARQUBE_TOKEN","-e","SONARQUBE_ORG",
    "sonarsource/sonarqube-mcp"],
  "env":{"SONARQUBE_TOKEN":"<YOUR_TOKEN>",
         "SONARQUBE_ORG":"<YOUR_ORG>"}}}}

[mcp_servers.sonarqube]
command = "docker"
args = ["run","--init","--pull=always","-i","--rm",
        "-e","SONARQUBE_TOKEN","-e","SONARQUBE_URL",
        "sonarsource/sonarqube-mcp"]
env = { SONARQUBE_TOKEN = "<YOUR_USER_TOKEN>",
        SONARQUBE_URL = "<YOUR_SERVER_URL>" }

Every other client

Each block below is the configuration for one client, with the file path and the key that client expects.

{"mcpServers":{"sonarqube":{"command":"docker",
  "args":["run","--init","--pull=always","-i","--rm","-e","SONARQUBE_TOKEN","-e","SONARQUBE_ORG","sonarsource/sonarqube-mcp"],
  "env":{"SONARQUBE_TOKEN":"<YOUR_TOKEN>","SONARQUBE_ORG":"<YOUR_ORG>"}}}}

{"mcpServers":{"sonarqube":{"command":"docker",
  "args":["run","--init","--pull=always","-i","--rm","-e","SONARQUBE_TOKEN","-e","SONARQUBE_ORG","sonarsource/sonarqube-mcp"],
  "env":{"SONARQUBE_TOKEN":"<YOUR_TOKEN>","SONARQUBE_ORG":"<YOUR_ORG>"}}}}

{"servers":{"sonarqube":{"command":"docker",
  "args":["run","--init","--pull=always","-i","--rm","-e","SONARQUBE_TOKEN","-e","SONARQUBE_ORG","sonarsource/sonarqube-mcp"],
  "env":{"SONARQUBE_TOKEN":"<YOUR_TOKEN>","SONARQUBE_ORG":"<YOUR_ORG>"}}}}

{"mcpServers":{"sonarqube":{"command":"docker",
  "args":["run","--init","--pull=always","-i","--rm","-e","SONARQUBE_TOKEN","-e","SONARQUBE_ORG","sonarsource/sonarqube-mcp"],
  "env":{"SONARQUBE_TOKEN":"<YOUR_TOKEN>","SONARQUBE_ORG":"<YOUR_ORG>"}}}}

docker run --init --pull=always -i --rm \
  -e SONARQUBE_TOKEN -e SONARQUBE_ORG \
  -e SONARQUBE_READ_ONLY=true \
  -e SONARQUBE_TOOLSETS=issues,quality-gates \
  sonarsource/sonarqube-mcp

docker run --init --pull=always -i --rm \
  -e SONARQUBE_TOKEN -e SONARQUBE_ORG \
  -e SONARQUBE_URL=https://sonarqube.us \
  sonarsource/sonarqube-mcp

Read-only access, permissions and security

• A read-only mode exists, and it is off by default

  SONARQUBE_READ_ONLY defaults to false, so issue and hotspot status changes are available on a bare connection.

• Server-level restrictions cannot be lifted per request

  In streamable HTTP mode a client may narrow the toolset or force read-only for its own request, but cannot widen beyond what the server was launched with.

• Two tools change triage state

  change_sonar_issue_status and change_security_hotspot_status alter how a finding is recorded.

• System tools expose the platform

  get_system_info and get_system_logs return SonarQube's own configuration and logs, which is more than a code-quality question needs.

Troubleshooting

A documented tool is missing

Its toolset is probably not enabled. SONARQUBE_TOOLSETS replaces the default selection, so naming one toolset disables the rest apart from projects.

Context tools return nothing

Vortex context and analysis tools need an organization entitlement and are documented as available only in stdio mode.

A status change is refused

SONARQUBE_READ_ONLY is set, or a client sent the read-only header for that request. A server-level restriction cannot be lifted per request.

A startup warning mentions deprecation

The cag and analysis toolset keys are deprecated in favor of the unified vortex key. The old keys still work and emit a warning.

SonarQube MCP server: Reference

What engineers report

The repository carries 631 stars, 94 forks and 37 releases since 30 April 2025. Its README is unusually specific about switches and entitlements, so the counts below come from it directly.

“The tool list this server advertises costs roughly 11,000 tokens, which every client pays on every conversation before a single tool is called.”

“I would like to track which tool was executed by which user and with what context or arguments. We need this for compliance and security reasons.”

Repository facts from the GitHub API, retrieved 24 August 2026. Tool inventory, toolsets and switches read from the repository README on the same date. The server is distributed only as a container image, and no container runtime was available in this environment, so no tools/list capture was made.

Should you connect an agent to SonarQube?

Yes. Best for a narrowed toolset with SONARQUBE_READ_ONLY=true, widened only where an agent is meant to triage.

Strongest fit

SonarQube teams who want an agent inside an existing quality workflow rather than a separate scanner.

Main advantage

One of the few servers here that treats context cost as configuration, with toolset selection and a read-only switch. GitHub pairs GITHUB_TOOLSETS with --read-only, Snyk documents --profile, and Rootly offers a slim tool profile.

Main weakness

Container-only distribution, so a machine without a container runtime cannot run it at all.

Operational risk

Read-only mode defaults to false, so triage-changing tools are available on a bare connection.

Frequently asked questions