# kanareika [Health: Active]

**Category:** 🔒 Security  
**Repository:** https://github.com/mikekosulin/kanareika  
**GitHub Stars:** 6  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/kanareika

## Description
Agentless, read-only security audit of Linux servers over SSH. Ranks what to patch first: CVEs with KEV and EPSS, SSH access, exposed ports, host health, with rollback-ready fixes. Terminal UI, web UI and an MCP server in one Go binary. No telemetry.

## Claude Desktop Quick Installation
No generated config for this server: it installs through its own setup steps.

The project's own setup steps (from https://github.com/mikekosulin/kanareika), copied verbatim. Run them in order; they are not an mcpServers config.

1. Install:

```sh
brew install mikekosulin/tap/kanareika
```

2. Install:

```sh
curl -fsSLO https://raw.githubusercontent.com/mikekosulin/kanareika/main/install.sh
```

3. Check the setup:

```sh
less install.sh
```

4. Install:

```sh
sh install.sh
```

5. Install:

```sh
cd "$(mktemp -d)"
```

6. Configure:

```sh
OS=$(uname -s | tr '[:upper:]' '[:lower:]')                 # linux, darwin
```

## Documentation

## What kanareika does

kanareika audits Linux servers over SSH without installing an agent on the target hosts. Its MCP mode gives AI agents access to the same audit data used by the terminal interface, web interface, and command-line scans.

The audit covers several areas:

- Operating-system packages, container images, JAR files, and lockfiles through the embedded Trivy scanner.
- Vulnerability prioritization using CISA KEV status, EPSS, exposure, and pre-authentication remote-code-execution risk.
- SSH configuration, password logins, keys, login history, attackers, open ports, firewalls, and external scan results.
- Host health signals such as kernel errors, disk status, memory, time synchronization, and pending reboots.

Findings are organized by priority from P1 through P4. Available fixes include verification and rollback steps, so an engineer or agent can review a proposed action before applying it.

## How it works

Run the Go binary's MCP entry point with `kanareika mcp`. The server reads organization configuration and scan data locally, while SSH connections provide access to the authorized Linux hosts. Organizations keep their settings in separate YAML files, which helps separate environments such as different clients or fleets.

Initial setup collects SSH users and keys, hosts, and consent to scan. The first scan may download vulnerability data of about 210 MB. Public vulnerability and reputation data can be downloaded from the listed upstream sources, while IP lookups use offline DB-IP Lite databases after they are available.

The same data can also be collected with `kanareika scan <ORG_ID>` and then consumed through the MCP interface. For isolated networks, databases can be exported on an internet-connected machine and imported before running an offline scan.

## Setup and configuration

Install the binary through Homebrew, the project's install script, or manually from the latest release archives. Linux and macOS archives support amd64 and arm64; Windows packages are available for amd64 and arm64.

Create an organization before scripted scans:

```bash
kanareika org new <ORG_ID>
kanareika scan <ORG_ID>
kanareika mcp
```

Hosts can come from an SSH configuration, a pasted `user@host` value, a host list, CSV, or Ansible inventory. The web interface listens on localhost by default, and remote web access uses a token. Changed host keys are refused, and scans require consent.

## Tools and capabilities

- Expose read-only audit results to AI agents through MCP.
- Rank vulnerabilities using KEV, EPSS, exposure, and exploitability signals.
- Inspect SSH access, open ports, firewalls, and external exposure.
- Review kernel, disk, memory, time-sync, and reboot health findings.
- Present fix actions with verification and rollback guidance.
- Reuse organization data across MCP, TUI, web, and command-line workflows.

## Limitations and notes

The audit reports discovered conditions but cannot guarantee complete coverage. Priorities and remediation steps are recommendations that should be reviewed before execution. The server is intended for systems that the operator is authorized to audit.

The documented scan and audit behavior is read-only; remediation becomes a task or recommendation rather than an automatic host change. Data and reports remain on the local machine, and telemetry is disabled, including telemetry in the embedded Trivy component. The vulnerability databases and other public datasets still need to be downloaded unless an offline bundle is imported.

The web interface binds to localhost by default. Organizations are isolated, but access to the local files, SSH credentials, and MCP process should still be controlled appropriately.

_Full upstream README: https://allmcps.com/mcp/kanareika/readme_