Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Description from the maker (original English)
Manage local KVM/libvirt VMs and Timeshift/Borg backups on Fedora: start, stop, clone, snapshot.
These three are never merged into one score: opinions, measurements and popularity are different things.
We have not written a plain-language summary for this one yet. Below are the facts we read from the registry.
Pick the AI app you use, then copy the settings.
MCP_AGENTS_LOG_DIRYou need
Which app do you use?
Pick your app above. We remember it, so next time the button says “Install in …”.
Open Claude Desktop → Settings → Developer → Edit Config, and paste this into claude_desktop_config.json
{
"mcpServers": {
"fedora-agents": {
"command": "npx",
"args": [
"-y",
"fedora-agents-mcp"
],
"env": {
"MCP_AGENTS_LOG_DIR": "your-value"
}
}
}
}Run this in your terminal
claude mcp add fedora-agents --env MCP_AGENTS_LOG_DIR=your-value -- npx -y fedora-agents-mcpFill in the empty values before running it.
Paste into .cursor/mcp.json in your project, or ~/.cursor/mcp.json
{
"mcpServers": {
"fedora-agents": {
"command": "npx",
"args": [
"-y",
"fedora-agents-mcp"
],
"env": {
"MCP_AGENTS_LOG_DIR": "your-value"
}
}
}
}Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.
Paste into .vscode/mcp.json in your project
{
"servers": {
"fedora-agents": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"fedora-agents-mcp"
],
"env": {
"MCP_AGENTS_LOG_DIR": "your-value"
}
}
}
}Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.
This one runs on your computer, and ChatGPT can only connect to servers on the internet. Pick another app.
Your choice stays on this device in a cookie; no account needed.
MCP_AGENTS_LOG_DIR envThese settings are generated from the maker’s registry entry. If the README says otherwise, follow the README.
Runs on your computer; the tool list appears in your AI app after you install it.
We check information anyone can see; we have not read the code. Checks that do not apply are left out of the count. · Checked Nov 7, 2026
The registry name was verified with the GitHub account amineutron, which owns the repo
We have not read this repo from GitHub yet
We have not read this repo from GitHub yet
This one runs on your computer; there is no internet endpoint
Runs on your computer and uses local settings instead
Runs on your computer; we can only read tool lists from internet endpoints
Compared with every active server in this category in the registry. Stars count only when the publisher owns the repo.
It passed 1 of 1 applicable basic checks. We have not read the code. Before adding a key or token, see what access the server asks for and use a key with only the permissions you need.
The registry does not list a required key, though some servers ask you to sign in on first use.
It runs on your computer. Your AI app starts it when needed, and data does not pass through the maker’s servers (unless it calls an online service itself).
The GitHub repo last changed –, and the latest version (1.3.0) reached the registry 42 days ago.
Apps that support MCP, such as Claude Desktop, Claude Code, Cursor and VS Code. You also need Node.js on your computer.
In the same category: Spec Driven Development, Remote Desktop Commander, Zendesk
Servers in the same category from other makers, sorted by stars.
Reading a card:★ GitHub starsiThree pillars: reviews · checks · usageiUpdated = latest version published to the registryi
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Remote Desktop Commander: use files, terminal and processes on your own computers from any AI app.
Zendesk MCP Pack — tickets, users, organizations via OAuth.
Copied from the maker’s repo as written. PopNx MCP has not reviewed or endorsed it. HTML, images and scripts are removed. Links lead off this site.
English summary. MCP server that gives an AI agent hands on KVM/libvirt virtual machines and Borg/Timeshift backups on Fedora. Each tool declares whether it needs sudo and whether it is destructive; the client must confirm destructive calls. Arguments are validated with Zod before any script runs. Install with npx fedora-agents-mcp (--help lists the settings) or npm ci && npm run build. Security policy and tool table: SECURITY.md.
Serveur MCP (Model Context Protocol) qui expose les agents VM-Controller et Backup-Manager via le protocole MCP. Permet a Claude Code de gerer les VMs KVM et les backups directement.
Un script virsh suffit à un humain qui lit la sortie et réagit. Un agent LLM (grand modèle de langage) qui pilote des machines a besoin d'autres garanties, et c'est ce que ce serveur ajoute autour des scripts :
src/logger.ts).vm_status, backup_status, backup_list, backup_verify) sont retentées, trois fois avec un délai croissant de 1, 2 puis 5 s ; une action qui modifie (démarrer, arrêter, restaurer, nettoyer) ne l'est jamais. Une erreur de validation ou de permission arrête tout de suite (src/utils/executor.ts).ErrorCode lisible par l'agent, au lieu d'un texte à interpréter (EXIT_CODE_MAP, src/config.ts).virsh, virt-clone ou qemu-img ; seuls les scripts système qui en ont besoin sont autorisés, un par un (voir Emplacement des scripts et sudoers).scripts/utils/tracking.sh) : l'agent vérifie l'avancement plus tard au lieu de bloquer.readOnlyHint, destructiveHint, idempotentHint) ; Lyra y lit la dangerosité et fait confirmer l'humain avant une action destructrice.Local-first et self-hosted : tout tourne sur la machine (on-prem), sans service tiers, avec un modèle local via Ollama dans Lyra ou n'importe quel client MCP.
Enregistree sur la machine reelle avec docs/demo/record.sh : un client MCP minimal (docs/demo/mcp_demo.py) demarre le serveur en stdio, liste les outils, appelle help puis vm_status (lecture seule). Les adresses IP sont remplacees par des adresses d'exemple.
fedora-agents/
scripts/ -- scripts bash embarques (autonomes, plus de dependance a fedora-setup) agents/vm-controller/ vm-start, vm-stop, vm-status, vm-exec, vm-copy, vm-snapshot, vm-destroy, vm-export, vm-import (+ common.sh) agents/backup-manager/ backup-create, backup-list, backup-restore, backup-verify, backup-clean, backup-status (+ common.sh) kvm/ kvm-clone, kvm-clone-system, kvm-snapshot, verify-vm-clone (+ helpers fix-nm-connection-vm, _fix-grub-vm) backup/ borg-backup, test-restore, get-borg-passphrase utils/tracking.sh reporting optionnel vers MCP Tracking (127.0.0.1:8765) config.env.example surcharges locales (KVM_IMAGES_DIR, VM_SSH_USER...) src/ index.ts -- point d'entree, enregistrement des outils MCP config.ts -- timeouts, permissions, codes d'erreur logger.ts -- logging JSON structure tools/ vm-controller.ts -- outils vm_start, vm_stop, vm_status, vm_exec, vm_copy, vm_snapshot, vm_verify, vm_clone, vm_clone_system, vm_destroy backup-manager.ts -- outils backup_create, backup_list, backup_restore, backup_verify, backup_clean, backup_status vm-portability.ts -- outils vm_export, vm_import utils/ executor.ts -- execution des scripts bash avec retry, timeout, gestion erreurs validation.ts -- schemas Zod pour tous les parametres d'outils
Liste tous les outils disponibles avec leurs descriptions.
| Outil | Script sous-jacent | Description |
|---|---|---|
vm_start | vm-controller/vm-start.sh | Demarre une VM, attend optionnellement SSH |
vm_stop | vm-controller/vm-stop.sh | Arrete une VM (proprement ou force) |
vm_destroy | vm-controller/vm-destroy.sh | Supprime definition + stockage d'une VM |
vm_status | vm-controller/vm-status.sh | Affiche l'etat d'une VM (ou liste toutes) |
vm_exec | vm-controller/vm-exec.sh | Execute une commande dans une VM via SSH |
vm_copy | vm-controller/vm-copy.sh | Copie des fichiers hote <-> VM via SCP |
vm_snapshot | vm-controller/vm-snapshot.sh | Gere les snapshots (create/list/restore/delete) |
vm_verify | kvm/verify-vm-clone.sh | Verifie qu'un clone est fidele au systeme hote |
vm_clone | kvm/kvm-clone.sh | Clone une VM existante (complet ou lie) |
vm_clone_system | kvm/kvm-clone-system.sh | Clone le systeme hote entier vers une VM |
| Outil | Script sous-jacent | Description |
|---|---|---|
backup_status | backup-manager/backup-status.sh | Dashboard global des backups |
backup_list | backup-manager/backup-list.sh | Liste les backups disponibles |
backup_create | backup-manager/backup-create.sh | Cree un backup (timeshift/borg/vm-snapshot/manual) |
backup_verify | backup-manager/backup-verify.sh | Verifie l'integrite des backups |
backup_restore | backup-manager/backup-restore.sh | Restaure un backup (destructif) |
backup_clean | backup-manager/backup-clean.sh | Applique les politiques de retention |
| Outil | Script sous-jacent | Description |
|---|---|---|
vm_export | vm-controller/vm-export.sh | Exporte une VM en archive .tar.gz sanitarisee |
vm_import | vm-controller/vm-import.sh | Importe une VM depuis une archive vm_export |
Ces scripts sont utilises en interne mais pas directement accessibles via MCP:
| Script | Role |
|---|---|
agents/vm-controller/common.sh | Fonctions communes (virsh, SSH, logging) |
agents/backup-manager/common.sh | Fonctions communes (borg, locks, notifications) |
kvm/kvm-snapshot.sh | Moteur de snapshots appele par vm-snapshot.sh |
kvm/fix-nm-connection-vm.sh, kvm/_fix-grub-vm.sh | Corrections post-clone (NetworkManager, grub BLS) |
backup/borg-backup.sh, backup/get-borg-passphrase.sh | Sauvegarde Borg ; passphrase via systemd-creds ou Bitwarden, jamais en clair |
backup/test-restore.sh | Test de restauration utilise par backup-verify --deep |
src/config.ts resout la racine des scripts dans cet ordre :
LYRA_SCRIPTS_DIR (variable d'environnement)/usr/local/lib/lyra/scripts : copie root:root 0755 installee par l'installeur Lyrascripts/ du depot (mode developpement)Les outils sans sudo (vm_status, vm_start, vm_stop, vm_exec, vm_copy, vm_snapshot, vm_export, vm_verify) pilotent libvirt directement : l'utilisateur doit etre membre du groupe libvirt, que polkit autorise sur qemu:///system :
sudo usermod -aG libvirt "$USER" # puis se reconnecter
virsh -c qemu:///system list --all # doit repondre sans sudo
Aucune regle sudoers ne vise virsh, virt-clone ou qemu-img : un NOPASSWD sur ces binaires equivaut a root (un domaine peut monter le disque de l'hote). Les outils qui demandent root (vm_destroy, vm_clone, vm_clone_system, vm_import et les backup_*) lancent leur script entier via sudo. Ces regles visent uniquement la copie systeme, script par script, jamais un glob sur un dossier inscriptible par l'utilisateur (sinon n'importe quel processus de son uid obtient root en y deposant un .sh) :
user ALL=(ALL) NOPASSWD: /usr/local/lib/lyra/scripts/kvm/kvm-clone.sh
user ALL=(ALL) NOPASSWD: /usr/local/lib/lyra/scripts/kvm/kvm-clone-system.sh ...
L'installeur Lyra genere ce fichier (/etc/sudoers.d/lyra) et le valide avec visudo -cf avant de l'activer. Installation manuelle :
sudo install -d -o root -g root -m 0755 /usr/local/lib/lyra/scripts
sudo cp -r scripts/. /usr/local/lib/lyra/scripts/ sudo chown -R root:root /usr/local/lib/lyra/scripts sudo find /usr/local/lib/lyra/scripts -type f -name '*.sh' -exec chmod 0755 {} +
| Variable | Rôle | Défaut |
|---|---|---|
SCRIPTS_DIR | dossier des scripts (agents/, kvm/) : copie root pour la production, scripts/ du dépôt pour les tests | copie root installée par Lyra, sinon scripts/ |
MCP_AGENTS_LOG_DIR | journaux JSON (serveur, erreurs, audit des appels sous sudo ou destructifs) | /var/log/mcp-agents s'il est inscriptible, sinon $XDG_STATE_HOME/mcp-agents (~/.local/state/mcp-agents) ; fichiers en 0600, jamais dans /tmp |
scripts/config.env | chemins KVM, Borg, Timeshift (voir config.env.example) | valeurs d'exemple |
VM_SSH_USER, VM_SSH_USERS | compte SSH des VMs : par defaut, puis par VM ("fedora-base=fedora ubuntu-base=ubuntu"), dans config.env ou ~/.config/vm-controller/config | utilisateur courant |
Le serveur est un composant de Lyra mais fonctionne seul : SCRIPTS_DIR=./scripts node dist/index.js démarre sans sudoers (les outils marqués requiresSudo échoueront alors proprement).
src/config.ts (TIMEOUTS)src/config.ts (TOOL_PERMISSIONS)src/config.ts (RETRY_CONFIG)scripts/config.env (copier depuis scripts/config.env.example)Les logs JSON structures sont ecrits dans:
/var/log/mcp-agents/ (si accessible en ecriture)~/.local/state/mcp-agents/ (fallback utilisateur)/tmp/mcp-agents-logs/ (fallback final)npx fedora-agents-mcp # depuis npm ; `npx fedora-agents-mcp --help` pour les reglages
Configuration Claude Desktop / Claude Code (mcpServers) :
{ "fedora-agents": { "command": "npx", "args": ["-y", "fedora-agents-mcp"] } }
Politique de securite et tableau des outils : SECURITY.md.
Paquets a installer sur l'hote (noms Fedora ; Debian/Ubuntu entre parentheses quand ils different) :
| Commande | Paquet | Outils concernes |
|---|
Only the beginning is shown; read the rest in the repo