Linux MCP Server for Claude
Run Linux system diagnostics from Claude — inspect CPU, memory, disk, network, services, processes, and journal logs — with the official Linux MCP Server from Red Hat's RHEL Lightspeed team, supporting both local and remote SSH-connected hosts.
Open the source and read safety notes before installing.
Safety notes
- The default `fixed` toolset is read-only — no changes are made to the system; only diagnostic information is retrieved.
- Set `LINUX_MCP_TOOLSET=run_script` to enable script execution — this allows arbitrary command execution on the host; only enable in trusted environments.
- For remote hosts, `LINUX_MCP_SSH_KEY_PATH` is used for key-based SSH authentication — ensure the key is protected with appropriate permissions.
Privacy notes
- System information including hostname, OS release, CPU model, active network connections, service names, process names, and journal log content are surfaced in Claude's context.
- Journal logs may contain sensitive application data — scope `LINUX_MCP_ALLOWED_LOG_PATHS` to limit which logs are readable.
Prerequisites
- A Linux system running locally or accessible via SSH.
- Python with `uvx` available.
- An MCP client such as Claude Code or Claude Desktop.
Schema details
- Install type
- cli
- Troubleshooting
- No
- Scope
- Source repo
- Estimated setup
- 5 minutes
- Difficulty
- beginner
Full copyable content
{
"mcpServers": {
"linux": {
"command": "uvx",
"args": ["linux-mcp-server"]
}
}
}About this resource
Overview
The Linux MCP Server is the official Model Context Protocol server from Red Hat's RHEL Lightspeed team, giving Claude read-only access to Linux system diagnostics. It covers system information, CPU, memory, disk usage, network connections, systemd services, journal logs, and running processes — for both local and SSH-connected remote hosts. Licensed under Apache-2.0.
The default fixed toolset is strictly read-only. An optional run_script toolset enables
arbitrary command execution when needed for advanced diagnostics.
Key capabilities
- System info — hostname, OS release, kernel version, uptime.
- CPU & memory — CPU model, core count, load averages, memory usage.
- Disk — filesystem mount point usage and capacity.
- Services — list all systemd units, get status of specific services.
- Journal logs — retrieve recent logs filtered by service or unit.
- Processes — list running processes sorted by CPU or memory.
- Network — active network connections and listening ports.
- Remote hosts — SSH key-based connections for remote system diagnostics.
Tools
| Tool | Purpose |
|---|---|
system_info |
Hostname, OS release, kernel, uptime |
cpu_info |
CPU model, core count, load averages |
memory_info |
Memory usage via free |
disk_usage |
Filesystem mount point usage |
list_services |
All systemctl service units |
service_status |
Status of a specific service |
service_logs |
Recent journal logs for a service |
journal_logs |
Filtered systemd journal queries |
list_processes |
Running processes sorted by CPU |
network_connections |
Active network connections |
Configuration options
| Env var | Default | Purpose |
|---|---|---|
LINUX_MCP_TOOLSET |
fixed |
fixed (read-only), run_script, or both |
LINUX_MCP_SSH_KEY_PATH |
— | SSH private key for remote hosts |
LINUX_MCP_KEY_PASSPHRASE |
— | SSH key passphrase |
LINUX_MCP_ALLOWED_LOG_PATHS |
— | Comma-separated allowed log file paths |
LINUX_MCP_COMMAND_TIMEOUT |
30 |
Command timeout in seconds |
LINUX_MCP_TRANSPORT |
stdio |
stdio, http, or streamable-http |
How it compares
| Server | Read-only mode | Remote SSH | systemd | Journal logs | Notes |
|---|---|---|---|---|---|
| Linux MCP | Yes (default) | Yes | Yes | Yes | Official, RHEL-backed |
| Desktop Commander | No | No | No | No | Full shell access |
| CLI MCP Server | No | No | No | No | General command runner |
Linux MCP is the only system diagnostics MCP with a read-only default mode, SSH remote support, and native systemd/journal integration from an enterprise Linux vendor.
Installation
Claude Code (local)
claude mcp add linux -- uvx linux-mcp-server
Claude Code (remote SSH host)
claude mcp add linux \
-e LINUX_MCP_SSH_KEY_PATH=/path/to/private-key \
-- uvx linux-mcp-server
Claude Desktop
{
"mcpServers": {
"linux": {
"command": "uvx",
"args": ["linux-mcp-server"]
}
}
}
Requirements
- A Linux system (local or remote via SSH).
- Python with
uvx(fromuv). - An MCP client (Claude Code or Claude Desktop).
Security
- The default
fixedtoolset is read-only — no system modifications are possible. - Only enable
run_scripttoolset if you need arbitrary command execution and trust the Claude session completely. - Restrict readable log paths with
LINUX_MCP_ALLOWED_LOG_PATHSin production environments.
Source Verification Notes
Verified on 2026-06-18:
- Official repository
rhel-lightspeed/linux-mcp-server(Apache-2.0) on PyPI aslinux-mcp-serverdocuments theuvxinstall, all diagnostic tools (system_info, cpu_info, memory_info, disk_usage, list_services, service_status, service_logs, journal_logs, list_processes, network_connections), thefixed/run_script/bothtoolset modes, SSH remote configuration, and optional env vars. - Claude Code MCP documentation at
code.claude.com/docs/en/mcpdescribes the stdio connector pattern used above.
Source citations
Add this badge to your README
How it compares
Linux MCP Server for Claude side by side with 3 alternatives on trust, install, platform support, and disclosed safety notes — all from reviewed registry metadata.
| Field | Linux MCP Server for Claude Run Linux system diagnostics from Claude — inspect CPU, memory, disk, network, services, processes, and journal logs — with the official Linux MCP Server from Red Hat's RHEL Lightspeed team, supporting both local and remote SSH-connected hosts. Open dossier | ConfigCat MCP Server for Claude Manage ConfigCat feature flags from Claude — create, update, and delete flags and targeting rules, manage environments, find and clean up stale flags, and audit change history — with the official ConfigCat MCP server and its 52 tools for the full ConfigCat Management API. Open dossier | OpenTelemetry MCP Server for Claude Analyze distributed traces and LLM observability data from Claude — search traces and spans, find errors, list services, analyze LLM token usage, identify slow LLM operations, and discover AI model usage patterns — with the OpenTelemetry MCP server supporting Jaeger, Grafana Tempo, and Traceloop backends. Open dossier | Radar MCP Server Built-in HTTP MCP server for Radar, the local-first Kubernetes UI that gives Claude token-optimized cluster topology, health, events, logs, audit, RBAC, GitOps, workload, and resource-management tools. Open dossier |
|---|---|---|---|---|
| Trust | ||||
| Install risk | Review first | Review first | Review first | Review first |
| Notes | Safety ✓ Privacy ✓ | Safety ✓ Privacy ✓ | Safety ✓ Privacy ✓ | Safety ✓ Privacy ✓ |
| Category | mcp | mcp | mcp | mcp |
| Source | source-backed | source-backed | source-backed | source-backed |
| Author | Red Hat RHEL Lightspeed | ConfigCat | Traceloop | Skyhook |
| Added | 2026-06-18 | 2026-06-18 | 2026-06-18 | 2026-06-06 |
| Platforms | Claude CodeCodexCursorClaude Desktop | Claude CodeClaude Desktop | Claude CodeClaude Desktop | Claude CodeClaude Desktop |
| Source repo | — | — | — | — |
| Safety notes | ✓The default `fixed` toolset is read-only — no changes are made to the system; only diagnostic information is retrieved. Set `LINUX_MCP_TOOLSET=run_script` to enable script execution — this allows arbitrary command execution on the host; only enable in trusted environments. For remote hosts, `LINUX_MCP_SSH_KEY_PATH` is used for key-based SSH authentication — ensure the key is protected with appropriate permissions. | ✓Tools can create, update, and delete feature flags, targeting rules, environments, and segments — changes affect live feature flag configuration. Use `list-staleflags` before deleting flags to identify zombie flags and avoid breaking active SDKs. | ✓All tools are read-only — the server queries trace data but does not modify your application or tracing backend. Trace data may contain sensitive information (request parameters, user IDs, SQL queries) — ensure Claude has appropriate access to this data. | ✓Radar MCP can inspect live Kubernetes resources, topology, warning events, logs, metrics, package inventory, audit findings, RBAC permissions, Helm releases, ArgoCD resources, and FluxCD resources. The MCP implementation includes mutating tools for applying YAML, patching resources, restarting or scaling workloads, triggering or suspending CronJobs, managing GitOps resources, and cordoning, uncordoning, or draining nodes. Mutating tools are RBAC-enforced and annotated as destructive-capable, but clients and operators should still require human review before running them. `apply_resource` with forced ownership can take server-side-apply fields from other managers such as Helm or GitOps controllers. `manage_node drain` can evict pods and affect live workloads. Radar provides flags and environment variables that disable MCP, disable exec/local terminal paths, or loosen localhost/origin protections; only relax those protections in trusted development contexts. |
| Privacy notes | ✓System information including hostname, OS release, CPU model, active network connections, service names, process names, and journal log content are surfaced in Claude's context. Journal logs may contain sensitive application data — scope `LINUX_MCP_ALLOWED_LOG_PATHS` to limit which logs are readable. | ✓Feature flag configurations, targeting rules, audience segments, SDK keys, and audit log entries from your ConfigCat account are surfaced in Claude's context. `CONFIGCAT_API_USER` and `CONFIGCAT_API_PASS` are Management API credentials — keep them in the MCP config env and never commit them to version control. | ✓Distributed trace content including service names, operation names, error messages, HTTP parameters, and LLM prompt/response metadata may be surfaced in Claude's context. For Traceloop backend, `BACKEND_API_KEY` is required and grants access to your Traceloop organization's trace data. | ✓Kubeconfig paths, cluster names, namespaces, resource names, manifests, labels, annotations, topology edges, metrics, events, logs, RBAC rules, Helm release data, GitOps metadata, and audit findings may be exposed to the MCP client and model provider. Radar's MCP docs state that Secret data is not exposed and environment/log values are redacted for known sensitive patterns, but metadata, key names, errors, and operational context can still be sensitive. Workload logs can contain credentials, customer data, request payloads, internal hostnames, incident details, or proprietary deployment context. In-cluster deployments without authentication make all MCP callers share the pod service account view, so restrict network access, enable auth, and scope RBAC before exposing shared endpoints. |
| Prerequisites |
|
|
|
|
| Install | | | | |
| Config | | | | |
| Citations | ||||
| Claim | Unclaimed | Unclaimed | Unclaimed | Unclaimed |
Signals
Loading live community signals…
A short, calm digest of reviewed Claude resources. Unsubscribe any time.