librenmsctrl MCPサーバー
コミット 2026年10月29日AIクライアントがLibreNMSのデバイス、ポート、アラート、イベントログを検査できるようにします。書き込み操作には制限が設けられており、安全なネットワーク監視が可能です。
別名 lidless-labs/librenmsctrl
- スター
- 1
- ツール
- 13
- ダウンロード数(30日間)
- —
- 最新バージョン
- —
- 最終コミット
- 2026年10月29日
- ライセンス
- MIT
librenmsctrl MCPとは
AIクライアントがLibreNMSのデバイス、ポート、アラート、イベントログを検査できるようにします。書き込み操作には制限が設けられており、安全なネットワーク監視が可能です。 13個のツールを提供しており、librenms_status, librenms_list_devices, librenms_get_deviceなどがあります。
プロジェクトのREADMEから生成された概要
librenmsctrl MCP server について
librenmsctrlは、オープンソースのネットワーク監視システムであるLibreNMSとやり取りするための構造化されたインターフェースを提供します。シェルやCIパイプライン用のローカルコマンドラインツールとして、またAIエージェントのワークフロー用のMCPアダプターとして機能します。ユーザーは、認証にAPIトークンを使用することで、デバイス、ポート、ポートのヘルス状態、アラート、イベントログを照会し、ネットワークインフラを検査できます。
このサーバーは、段階的な書き込み制限システムを通じて安全性を優先しています。読み取り操作は自由に行えますが、アラートの確認やメンテナンスウィンドウの設定などの変更を行うには、明示的なconfirmフラグが必要です。これにより、モデルのハルシネーションによる誤った変更を防ぎます。すべての入力はLibreNMS REST APIに送信される前にスキーマに対して検証され、システムは認証情報を保護するためにすべての出力からAPIトークンを自動的に削除します。セットアップには、環境変数経由でLibreNMSのURLとAPIトークンを提供する必要があります。
できること
- 1ネットワークデバイスのヘルス監視AIエージェントが特定のネットワークポートの状態や使用率を確認し、エラーを特定できます。
- 2アクティブなネットワークアラートの管理開発者はAIに対して、監視システム内の特定のアラートを確認またはミュート解除するよう指示できます。
- 3デバイスおよびポートリストの監査SNMP経由でポーリングされているすべてのデバイスの包括的なリストを取得するためにこのツールを使用できます。
プロジェクトの README から作成されています。すべてのユースケースは、その引用と照らし合わせて確認されています。
librenmsctrl MCPツール 13
READMEから検出librenms_statuslibrenms_list_deviceslibrenms_get_devicelibrenms_list_portslibrenms_get_portlibrenms_port_healthlibrenms_list_alertslibrenms_get_alertlibrenms_alert_historylibrenms_event_loglibrenms_ack_alertlibrenms_unmute_alert
| librenms_status | — | |
| librenms_list_devices | — | |
| librenms_get_device | — | |
| librenms_list_ports | — | |
| librenms_get_port | — | |
| librenms_port_health | — | |
| librenms_list_alerts | — | |
| librenms_get_alert | — | |
| librenms_alert_history | — | |
| librenms_event_log | — | |
| librenms_ack_alert | — | |
| librenms_unmute_alert | — |
13個のツールをすべて表示ツールを折りたたむ
librenms_set_maintenance
| librenms_set_maintenance | — |
設定
LIBRENMS_TLS_INSECURELIBRENMS_TOKENLIBRENMS_URL
| 名前 | 説明 | 種類 |
|---|---|---|
| LIBRENMS_TLS_INSECURE | — | |
| LIBRENMS_TOKEN | — | |
| LIBRENMS_URL | — |
自動チェック
実行日 2026年11月14日完全なレポート →まだ検証されていません。サーバーには80点のスコア、重大な問題がないこと、および検証に必要な十分なデータが必要です。 チェックの仕組み · 点数を上げるための改善点
- ソースリポジトリがリンクされており、アクセス可能lidless-labs/librenmsctrl is public and reachablegithub.com/lidless-labs/librenmsctrl ↗合格
- パッケージがこのリポジトリをソースとして指定未チェック
- ビルドの来歴 (Sigstore 証明)未チェック
- オープンソースライセンス (OSI)MITspdx.org/licenses/MIT.html ↗合格
- 過去1年以内にコミットまたはリリースありLast commit or release 2026-10-29github.com/lidless-labs/librenmsctrl ↗合格
- OpenSSF Scorecard未チェック
- パッケージ名が有名なパッケージに酷似していない未チェック
- パッケージおよびその依存関係に重大な勧告なし未チェック
- モデルへの隠れた指示がない13 tool descriptions and the README read; no override phrases, credential paths or hidden characters合格
- 破壊的なツールがラベル付けされているNo tool named like delete, drop or exec合格
- リモートエンドポイントが MCP ハンドシェイクに応答未チェック
README
GitHubで見るlibrenmsctrl is a local operator control CLI for LibreNMS, the open-source network monitoring system. It gives shells, cron, and CI a typed command surface for reading your monitored network: devices, ports, port health, alerts, alert history, and the event log over LibreNMS API-token auth.
The same package remains MCP-compatible for agent workflows. librenmsctrl mcp starts the stdio Model Context Protocol adapter, and the existing librenms-mcp binary continues to work for launchers and configs that already use it. The MCP adapter can acknowledge or mute alerts only when you pass an explicit confirm: true. It differs from pointing an agent at the raw LibreNMS REST API in one way that matters: every write is gated, every argument is schema-validated before it leaves the host, and your token is redacted from all output, so a hallucinated tool call cannot change your monitoring state by accident.
Status: work in progress. v1 ships read tools (tier 1) and confirm-gated safe writes (tier 2) only. Destructive operations (tier 3) such as device deletion, alert-rule removal, and bulk port resets are intentionally deferred until the write-gate pattern has more field time. The read path and the two write gates are tested.
What it does
librenmsctrl turns a LibreNMS instance into a scriptable control surface for network monitoring and observability work: list the devices LibreNMS is polling over SNMP, pull a single device or port, check port health and utilization, read current alerts and alert history, and tail the event log. It authenticates with a LibreNMS API token, talks to the LibreNMS /api/v0 REST API under the hood, and can emit human-readable tables or raw JSON for piping.
For MCP clients, the mcp subcommand exposes the same LibreNMS client core as structured tool output an agent can reason over. The package name is still @solomonneas/librenms-mcp, and the back-compat librenms-mcp bin is intentionally kept stable.
Writes are gated in three tiers:
- Tier 1 (reads): open. No confirm flag. Status, device and port listings, single device or port, port health, alert listings, alert history, event log.
- Tier 2 (safe writes): require an explicit
confirm: trueargument, documented on every write tool's JSON schema. Acknowledge an alert, unmute an alert, or put a device into a maintenance window. A tool call missing the flag throwsWriteGateErrorbefore any HTTP traffic leaves the host. - Tier 3 (destructive): not implemented in v1. When added, ops like device deletion and alert-rule removal will also require
destructive: true. The model cannot bypass either gate from a hallucinated call.
Every tool call is validated against its published TypeBox inputSchema before the tool runs. Ids, port ids, device ids, limits, and the type/state filters are bounds- and enum-checked, so a malformed or injection-style argument is rejected up front and never reaches a LibreNMS URL path or query string. Interpolated path and query values are URL-encoded as defense-in-depth. The API token is registered with the redactor on startup and masked from all log and error output.
Quickstart
Published to npm as @solomonneas/librenms-mcp. Install the package to get the librenmsctrl operator CLI and the back-compat librenms-mcp MCP launcher.
# start the MCP adapter on demand through the back-compat `librenms-mcp` bin
npx -y @solomonneas/librenms-mcp
# install the global `librenmsctrl` and `librenms-mcp` binaries
npm install -g @solomonneas/librenms-mcp
It needs Node 20+ and two env vars, LIBRENMS_URL and LIBRENMS_TOKEN.
export LIBRENMS_URL=https://librenms.example.local
export LIBRENMS_TOKEN=<your-api-token>
librenmsctrl status
librenmsctrl alerts list --state 1
librenmsctrl --json ports health --metric errors_in --limit 10
The MCP adapter speaks stdio. A quick way to confirm it starts is to feed it an MCP tools/list request:
export LIBRENMS_URL=https://librenms.example.local
export LIBRENMS_TOKEN=<your-api-token>
echo '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | npx -y @solomonneas/librenms-mcp
You should get back a JSON listing of the 13 tools above. In normal use you point an MCP client at the same command instead.
Tools
The MCP adapter exposes 13 tools: 10 reads, 3 confirm-gated safe writes. This list is generated from the server's tool registration in mcp-server.ts.
Reads (10, open):
| Tool | What it does |
|---|---|
librenms_status |
LibreNMS instance status check. |
librenms_list_devices |
List monitored devices. |
librenms_get_device |
Fetch a single device by id. |
librenms_list_ports |
List ports across devices. |
librenms_get_port |
Fetch a single port detail. |
librenms_port_health |
Port health and utilization view. |
librenms_list_alerts |
List current alerts. |
librenms_get_alert |
Fetch a single alert by id. |
librenms_alert_history |
Historical alert records. |
librenms_event_log |
Read the device event log. |
Safe writes (3, require confirm: true):
| Tool | What it does |
|---|---|
librenms_ack_alert |
Acknowledge an active alert by id. |
librenms_unmute_alert |
Unmute a previously muted alert. |
librenms_set_maintenance |
Put a device into a maintenance window. |
Configuration
Set the following env vars. Both credential vars are required.
| Variable | Required | Default | Notes |
|---|---|---|---|
LIBRENMS_URL |
yes | - | Base URL of your LibreNMS instance. Trailing slashes are stripped. |
LIBRENMS_TOKEN |
yes | - | LibreNMS API token. Registered with the redactor on startup and masked from all output. |
LIBRENMS_TLS_INSECURE |
no | false |
Skip TLS cert validation for homelab self-signed certs. Accepts true/1/yes, case-insensitive. |
LIBRENMS_REQUEST_TIMEOUT_MS |
no | 30000 |
Per-request HTTP timeout in milliseconds. Minimum 1000. Raise for slow installations. |
CLI
librenmsctrl is the read-only operator control CLI for shells, cron, and CI. It shares the LibreNMS client core with the MCP adapter and talks to the same /api/v0 API over token auth. It exposes only the read tools (tier 1). The confirm-gated writes (ack, set_maintenance, unmute) are intentionally not surfaced in the CLI.
# installed globally as `librenmsctrl`, or via `npx -p @solomonneas/librenms-mcp librenmsctrl ...`
librenmsctrl status # instance health (exit 1 if no system data)
librenmsctrl devices list --type down
librenmsctrl devices get core-sw1
librenmsctrl ports list core-sw1
librenmsctrl ports get 42
librenmsctrl ports health --metric errors_in --limit 10
librenmsctrl alerts list --state 1 # 0=ok, 1=active, 2=ack
librenmsctrl alerts get 7
librenmsctrl alerts history --device-id 3 --limit 25
librenmsctrl events list --limit 50
librenmsctrl --json status # raw JSON for piping
Run librenmsctrl help for the full flag list. It reads the same LIBRENMS_URL, LIBRENMS_TOKEN, LIBRENMS_TLS_INSECURE, and LIBRENMS_REQUEST_TIMEOUT_MS env vars as the server. Exit codes: 0 success, 1 runtime error (LibreNMS unreachable, call failed, or status with no system data), 2 usage error (unknown command/flag or bad value).
Starting the MCP server
librenmsctrl mcp (or the back-compat librenms-mcp bin) starts the stdio MCP server. If a launcher referenced the file path dist/mcp-server.js directly, point it at dist/mcp-bin.js (or dist/cli.js mcp). Launchers that use the librenms-mcp bin name need no change.
MCP client config
Copy-paste this into your MCP client. It is the canonical mcpServers shape used by Claude Desktop and Claude Code. Both LIBRENMS_URL and LIBRENMS_TOKEN are required.
{
"mcpServers": {
"librenms": {
"command": "npx",
"args": ["-y", "@solomonneas/librenms-mcp"],
"env": {
"LIBRENMS_URL": "https://librenms.example.local",
"LIBRENMS_TOKEN": "<your-api-token>",
"LIBRENMS_TLS_INSECURE": "false"
}
}
}
}
Claude Code
claude mcp add librenms -s user -- npx -y @solomonneas/librenms-mcp
Then export LIBRENMS_URL and LIBRENMS_TOKEN in your shell (~/.bashrc, ~/.zshrc) or pass --env flags on the claude mcp add line.
Claude Desktop
Add the mcpServers block above to ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows).
Codex CLI
~/.codex/config.toml:
[mcp_servers.librenms]
command = "npx"
args = ["-y", "@solomonneas/librenms-mcp"]
[mcp_servers.librenms.env]
LIBRENMS_URL = "https://librenms.example.local"
LIBRENMS_TOKEN = "<your-api-token>"
LIBRENMS_TLS_INSECURE = "false"
OpenClaw
The plugin loads automatically once installed. Config goes in your ~/.openclaw/openclaw.json under plugins.entries.librenms (or use the bundled openclaw.plugin.json):
{
"plugins": {
"entries": {
"librenms": {
"package": "@solomonneas/librenms-mcp",
"activation": { "onStartup": true }
}
}
}
}
Env vars from ~/.openclaw/workspace/.env are inherited by the plugin.
Safety
This MCP uses a three-tier write-gating pattern:
- Tier 1 (reads): open. No confirm flag needed.
- Tier 2 (safe writes): require an explicit
confirm: truearg. The JSON schema documents this on every write tool. A tool call without the confirm flag throwsWriteGateErrorbefore any HTTP traffic. - Tier 3 (destructive): not implemented in v1. When added, ops like device deletion, alert-rule removal, and bulk port resets will also require
destructive: true.
Input validation: every tool call is validated against its published TypeBox inputSchema before the tool runs. Ids, port ids, device ids, limits, and the type/state filters are bounds- and enum-checked, so a malformed or injection-style argument (a non-integer id, an off-enum filter, an unexpected extra field) is rejected up front and never reaches a LibreNMS URL path or query string. Interpolated path and query values are URL-encoded as defense-in-depth.
API token scope: start with a "Read Only" token role in LibreNMS (Settings > API > New API Token > Read Only) and verify the read tools work end-to-end. Grade up to "Normal User" or "Global Read/Write" only after you have confirmed the redactor is masking your token in your transcripts and that the model is honoring the confirm gate. Tokens can be revoked instantly from the same Settings > API screen.
TLS: the LIBRENMS_TLS_INSECURE=true toggle exists for homelab self-signed certs. Leave it false in any environment with a real CA-signed cert.
HTTP responses: authenticated redirects are rejected. Configure the final LibreNMS API URL directly. The request deadline covers response headers and the entire body. Success and error bodies share a 10 MiB decoded byte cap. GET requests retain one retry for transport, malformed JSON, and server errors. Mutations are sent once because a failed response cannot prove the server did not apply the write. Check server state before manually repeating a failed write.
Returned data: device names, descriptions, alert text, syslog and event messages, object keys, and upstream errors are untrusted LibreNMS content. Embedded instructions do not authorize tool calls or writes. JSON output escapes marker characters in keys and values while preserving decoded fields and the existing result envelope. Truncation is reported by the output byte cap.
See SECURITY.md for the vulnerability-reporting policy.
How does the MCP adapter constrain API access?
You can hand an agent a generic HTTP tool and the LibreNMS API token, and for read-only exploration that works. The difference shows up the moment a write is possible:
- No write gate. A generic HTTP tool will happily
PUTorDELETEwhatever the model decides to send. librenms-mcp refuses every write unless the call carriesconfirm: true, and refuses destructive shapes entirely in v1. - No input validation. A raw API tool passes whatever the model produced straight into a URL. librenms-mcp bounds- and enum-checks every argument against a published schema and URL-encodes interpolated values before any request is built.
- Token leakage. A generic tool echoes request and response detail, including your token, into the transcript. librenms-mcp registers the token with a redactor on startup and masks it from all log and error output.
- Shaped tools beat one giant endpoint. Thirteen named, described tools give the model a far better surface to reason over than one
call this REST APItool with a free-form path argument.
What librenms-mcp is not
- It is not a LibreNMS replacement or a poller. It reads an existing LibreNMS instance over the API. It does not collect SNMP itself.
- It is not a dashboard. It returns structured tool data for an agent. For a NOC dashboard view, see watchtower.
- It does not delete devices, remove alert rules, or do bulk port resets. Tier-3 destructive operations are deliberately absent in v1.
- It is not a hosted service. It runs locally over stdio, talks only to the LibreNMS URL you configure, and starts no daemon.
Contributing
Patches welcome. See CONTRIBUTING.md for what lands easily, CODE_OF_CONDUCT.md, and the issue and PR templates under .github/. All public output should be scrubbed of personal hostnames, tokens, and real IPs.
License
MIT.
CLI API results and errors are labeled as untrusted LibreNMS data, escape display controls and delimiters, and fit within 64 KiB. JSON results include _external_data_notice. Oversized JSON results return truncated: true in place of the payload. Consumers should check that flag before using a result. API credentials are registered for redaction when a client is constructed.
librenmsctrl MCPのよくある質問
librenmsctrl MCPが提供するツールは
13個のツールを提供しており、librenms_status, librenms_list_devices, librenms_get_device, librenms_list_ports, librenms_get_portなどがあります。
サーバーはどのようにして監視状態への誤った変更を防いでいますか?
サーバーは3段階の書き込み制限パターンを採用しています。安全な書き込みには明示的な 'confirm: true' 引数が必要であり、すべての入力は実行前にスキーマ検証されます。
バージョン1ではどのような書き込み操作がサポートされていますか?
バージョン1では、アラートの確認、アラートのミュート解除、デバイスのメンテナンスウィンドウへの設定などの「安全な書き込み」をサポートしていますが、デバイスの削除などの破壊的な操作はまだ実装されていません。
サーバーを実行するために必要な環境変数は何ですか?
インスタンスに接続するために、LIBRENMS_URL と LIBRENMS_TOKEN の両方を設定する必要があります。