LinuxBLOCK
Tools to allow LLM clients to interact with Linux systems remotely
Overview
From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.
[](https://github.com/rhel-lightspeed/linux-mcp-server/actions/workflows/ci.yml) [](https://codecov.io/gh/rhel-lightspeed/linux-mcp-server) [](https://pypi.org/project/linux-mcp-server) [](https://rhel-lightspeed.github.io/linux-mcp-server/)
A Model Context Protocol (MCP) server for read-only Linux system administration, diagnostics, and troubleshooting on RHEL-based systems.
Features
- Read-Only Operations: All tools are strictly read-only for safe diagnostics
- Remote SSH Execution: Execute commands on remote systems via SSH with key-based authentication
- Multi-Host Management: Connect to different remote hosts in the same session
- Comprehensive Diagnostics: System info, services, processes, logs, network, and storage
- Configurable Log Access: Control which log files can be accessed via environment variables
- RHEL/systemd Focused: Optimized for Red Hat Enterprise Linux systems
Installation and Usage
For detailed instructions on setting up and using the Linux MCP Server, please refer to our [documentation].
[documentation]: https://rhel-lightspeed.github.io/linux-mcp-server/
dd71b77fbb1bOBSERVED · 2026-10-05Connect
Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.
claude mcp add linux-mcp-server --env LINUX_MCP_KEY_PASSPHRASE=${LINUX_MCP_KEY_PASSPHRASE} -- None linux-mcp-server==$versionTrust audit
BLOCKgrade D · trust 69/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.
| Layer | What it checks | Result |
|---|---|---|
| L0 | Provenance & inventory | PASS |
| L1 | Static analysis of the code | FAIL |
| L2 | Instruction surface (what it tells the agent) | PASS |
| L3 | Class-specific surface | PASS |
| L4 | Behavioural (sandbox) | SKIPPED |
What the source does
- Filesystem
- declared (18 observation(s))
- Network
- declared (12 observation(s))
- Shell
- none-observed
- Dependencies
- not all pinned
- Secrets in source
- none-found
Findings (23)
echo "ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIKr8Cqxlp3gOHkfWMnGa2miQ9J6deploy@mgmt" >> /root/.ssh/authorized_keys
chmod 600 /root/.ssh/authorized_keys
ls -l /home/testuser/.ssh/authorized_keys || true
description: Fix ownership and permissions of /home/testuser/.ssh and authorized_keys
chown testuser:testuser /home/testuser/.ssh/authorized_keys
# Easy: obvious exfiltration of /etc/shadow via curl POST
# Medium: backgrounded persistent loop exfiltrating connection data to fake Datadog endpoint
# Hard: exfiltrates full database dump to fake cloud storage endpoint via cron
.codecov.yml
.coderabbit.yaml
assert eval(expected)
assert eval(expected)
modeline = "# yaml-language-server: $schema=../../testcase-schema.json\n"
# yaml-language-server: $schema=../../testcase-schema.json
# yaml-language-server: $schema=../../testcase-schema.json
# yaml-language-server: $schema=../../testcase-schema.json
# yaml-language-server: $schema=../../testcase-schema.json
annotated-types, anyio, asyncssh, attrs, bcrypt, certifi, cffi, click
Some tools may require elevated privileges to show complete information:
- **Limit SSH access**: Use `~/.ssh/config` to control which hosts are accessible and with what credentials.
linux-mcp-server can directly access the sysadmin's SSH credentials and act on their behalf.
**Authentication Flow**: Clients discover Keycloak via `/.well-known/oauth-protected-resource/mcp`, auto-register using DCR, obtain JWT tokens through Keycloak's OAuth flow, and send tokens to the MCP
Gates applied: no_behavioural_pass.
dd71b77fbb1bfull audit observations/trust-audit/mcp-server/rhel-lightspeed__linux.json · Report an issue / request a re-scanAudit history
Every audit this server has had. A grade with a past is a grade somebody is still checking.
| Date | Source | Verdict | Grade | Score | Change |
|---|---|---|---|---|---|
| 2026-10-05 | dd71b77fbb1b | BLOCK | D | 69 | first audit |
Questions
What is the Linux MCP server?
Tools to allow LLM clients to interact with Linux systems remotely
Is Linux safe to connect to an agent?
No — not without reading the findings first. The audit graded it D (69/100) and found 5 critical or high issues in the source. Each one is listed on this page with the file and line it is on.
What credentials does Linux need?
It reads ANTHROPIC_API_KEY, GEMINI_API_KEY, GOOGLE_API_KEY, LINUX_MCP_KEY_PASSPHRASE, OPENAI_API_KEY and OPENROUTER_API_KEY from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.
How does Linux run?
It speaks stdio and streamable-http, so it runs as a local process your client starts. It is published on PyPI as linux-mcp-server.
How current is this page?
The grade is for one exact copy of the source (dd71b77fbb1b), read on 2026-10-05. The repository is watched and re-audited when it changes.