wazuh-mcp

Active

Your Wazuh console answers one query at a time, and only if you know how to ask. wazuh-mcp hands the whole SIEM to your agent: alerts, agents, CVEs, rules, and CIS posture, answered in plain language over stdio.

GitHub repo → npm: wazuh-mcp →

28 tools3 resources3 promptsread-onlyNode 20+MIT

See it work

one question, the tools it calls, the answer

representative session · claude + wazuh-mcp

Any critical alerts on the DMZ hosts in the last hour, and is anything misconfigured?

search_alerts{ level: ">=10", group: "dmz", range: "1h" }

3 alerts 5710 sshd: 47 authentication failures dmz-web-01 (192.0.2.24) level 10 02:14Z 92053 Suricata ET SCAN nmap probe dmz-fw-01 level 12 03:02Z 5402 successful sudo to root dmz-web-01 level 12 03:05Z

get_sca_checks{ agent: "dmz-web-01", result: "failed" }

CIS 5.2.5 SSH MaxAuthTries not set fix: set MaxAuthTries 4 in /etc/ssh/sshd_config

dmz-web-01 is the one to look at. 47 SSH failures, then a successful root sudo three minutes after an nmap scan from dmz-fw-01. Its SSH config also fails CIS 5.2.5, which is what let the brute force run. I would isolate dmz-web-01 and cap auth tries.

Representative session. Illustrative values, RFC 5737 addresses. Your data never leaves the box.

What it does

Point any MCP client at wazuh-mcp and your Wazuh SIEM/XDR becomes tools the agent can call: enumerate agents, search alerts against the Indexer, review the CVE backlog, read detection rules and decoders, and run CIS and File Integrity posture checks. It is read-only and speaks stdio, so there is no new network surface and nothing to write back into Wazuh. Compliance mappings (PCI-DSS, HIPAA, NIST 800-53, MITRE ATT&CK) ride along on every rule lookup.

Capabilities

28 tools across the deployment

Agents & inventory9 tools

Enumerate agents and pull Syscollector OS, packages, processes, ports, and hotfixes.

Alerts & vulnerabilities5 tools

Search alerts by time, level, agent, and rule via the Indexer. Query the CVE inventory by severity and package.

Detection content4 tools

Browse rules and decoders with level/group filters and full compliance mappings.

Posture & integrity4 tools

SCA policies and per-check remediation, File Integrity Monitoring, and rootcheck findings.

Manager & diagnostics6 tools

Manager logs and config (secrets redacted), agent groups, version, and a sanitized connection check.

Full reference on GitHub →

Set it up

npm i -g wazuh-mcp # or: npx wazuh-mcp

Configure with WAZUH_URL, WAZUH_USERNAME, WAZUH_PASSWORD. Add WAZUH_INDEXER_URL to enable alert and vulnerability search.

MCP client config (Claude Desktop, Claude Code, OpenClaw, Codex)

{
  "mcpServers": {
    "wazuh": {
      "command": "wazuh-mcp",
      "env": {
        "WAZUH_URL": "https://wazuh.example.net:55000",
        "WAZUH_USERNAME": "wazuh-wui",
        "WAZUH_PASSWORD": "...",
        "WAZUH_INDEXER_URL": "https://wazuh.example.net:9200"
      }
    }
  }
}

[ shield ]

Built for pointing an LLM at security data. Sensitive fields (IPs, full logs, hashes, command lines, secrets) are minimized by default and returned only on explicit opt-in. Alert content that originated on a monitored endpoint is wrapped in untrusted-data markers, so a malicious log line cannot hijack the agent reading it.

MCP tools

resources

prompts

writes

MIT

license