managelm — плагин для Claude Code: серверы и установка

<p align="center">
  <a href="https://www.managelm.com">
    <img src="https://www.managelm.com/assets/ManageLM.png" alt="ManageLM" height="50">
  </a>
</p>

<h3 align="center">Claude Code Extension</h3>

<p align="center">
  Manage Linux &amp; Windows servers directly from Claude using natural language.
</p>

<p align="center">
  <a href="LICENSE"><img src="https://img.shields.io/badge/license-Apache--2.0-blue" alt="License"></a>
  <a href="https://www.managelm.com"><img src="https://img.shields.io/badge/website-managelm.com-cyan" alt="Website"></a>
  <a href="https://www.managelm.com/plugins/claude.html"><img src="https://img.shields.io/badge/docs-full%20documentation-green" alt="Docs"></a>
  <a href="https://github.com/managelm/claude-extension/releases"><img src="https://img.shields.io/github/v/release/managelm/claude-extension" alt="Release"></a>
</p>

<p align="center">
  <img src="assets/screenshot.png" alt="Claude managing a Linux server — creating users, managing SSH keys" width="600">
</p>

---

Ask Claude to check system status, manage packages, configure services, transfer files, run security audits, and more — across one server or an entire fleet. The extension connects Claude to your infrastructure through MCP (Model Context Protocol), a secure cloud portal, and lightweight agents running on your servers.

## Features

- **Natural language** — describe tasks in plain English; Claude picks the right skill and parameters
- **33 built-in skills** — packages, services, firewall, web servers, databases, containers and Kubernetes, certificates, VPN, Active Directory, and more, across Linux and Windows
- **Multi-server targeting** — run on a single server, a group, or broadcast to all agents
- **File transfers** — upload and download files (text and binary) between local and remote
- **Security audits** — run audits, search findings across your fleet, remediate issues
- **Task history & revert** — review what changed, inspect diffs, revert any task
- **Cross-infrastructure search** — find agents by health/OS, search inventory, security findings and threats, user activity, SSH keys, sudo rules, certificates, monitors, backups, scheduled tasks, credentials, keystore, and hosting resources
- **Zero inbound ports** — agents connect outbound only; no SSH, no open ports on your servers

## Quick Start

### 1. Install

As a Claude Code plugin (Claude Code 2.0.12+), which also brings the setup and troubleshooting skills:

```bash
claude plugin marketplace add managelm/claude-extension
claude plugin install managelm@managelm
```

Or add only the MCP server:

```bash
claude mcp add --transport http managelm https://app.managelm.com/mcp
```

Or add it to a project's `.mcp.json`:

```json
{
  "mcpServers": {
    "managelm": {
      "type": "http",
      "url": "https://app.managelm.com/mcp"
    }
  }
}
```

### 2. Authenticate

ManageLM uses **OAuth 2.0 with Dynamic Client Registration** (RFC 7591). On first connection, Claude self-registers as an OAuth client and opens your browser to authorize access — no client ID or secret to copy. All you need is a [ManageLM account](https://app.managelm.com/register) (free for up to 10 agents).

### 3. Use it

```
> Check disk usage on web-prod-01

> Install nginx on all servers in the staging group

> Upload my local config.yml to /etc/myapp/config.yml on db-primary

> Run a security audit on all production servers

> Which servers have CPU usage above 80%?

> Create a user deploy with my SSH key on web-prod-01
```

## Architecture

```
Claude ── MCP ──> ManageLM Portal ── WebSocket ──> Agent on Server
                  (cloud control      (outbound       (your LLM,
                   plane)              only)            skill exec)
```

Every task dispatched to an agent is cryptographically signed (Ed25519). Each agent calls the LLM you configure for it: with a local model, your data never leaves your infrastructure.

## Available Skills

Each skill imported into your account from the portal's catalog becomes one tool; a new account starts with `base` only. An account set to No LLM gets no skill tools.

| Skill | Description |
|-------|-------------|
| `base` | Read-only utilities — files, system info, resource usage, diagnostics (always assigned) |
| `system` | Hostname, timezone, locale, kernel modules, sysctls, bootloader |
| `packages` | Install, update, and remove packages and system updates |
| `services` | Services, processes, scheduled tasks, systemd units and timers |
| `users` | Accounts, groups, passwords, SSH keys, sudo |
| `network` | Adapters, IP addresses, routing, bridges, traffic shaping |
| `firewall` | Firewall rules, profiles, IP sets, access policies |
| `security` | fail2ban, SSH hardening, SELinux/AppArmor, auditd, antivirus |
| `files` | Read, write, upload, download files (text and binary) |
| `storage` | Disks, volumes, LVM, ZFS, Btrfs, LUKS, quotas |
| `containers` | Docker/Podman containers, Compose, Kubernetes, Helm |
| `webserver` | Web server sites, virtual hosts, HTTPS, modules |
| `database` | MySQL/MariaDB, PostgreSQL, SQLite: queries, backups, restores |
| `certificates` | SSL/TLS certificates, certificate stores, HTTPS bindings |
| `activedirectory` | Active Directory users, groups, OUs, group policies (Windows) |
| ...and more | Automation, backup, developer tools, DNS, email, file sharing, LDAP, LLM servers, logs, message queues, monitoring, NoSQL, proxy, RDP, registry, virtualization, VPN, web apps |

## Discovery & Management Tools

Beyond the skills, the portal gives Claude 33 built-in tools. A few only appear when they apply: `search_pki`, `search_monitors`, `search_backups`, `search_credentials` and `search_keystore` follow the optional features enabled on your account; the hosting tools appear once a hosting connector exists, and `cloud_action` only once a hosting connector has write mode on; on an account set to No LLM, `get_agent_skills`, `list_available_skills`, `answer_task` and `search_schedules` are hidden.

| Tool | Description |
|------|-------------|
| `search_agents` | List servers (status, OS, health, groups, site) or filter by CPU/memory/disk, OS, status, group, site |
| `get_agent_info` | Detailed info for a single server |
| `get_agent_skills` | Skills assigned to a server, and those that could be |
| `list_available_skills` | Catalog skills not yet imported into your account |
| `get_account_info` | Plan, usage limits, groups and sites |
| `list_team_members` | Team members with roles, permissions and SSH keys |
| `search_inventory` | Search packages, services, containers across the fleet |
| `search_security` | Search audit findings, pentest findings and threat alerts, with a time range |
| `search_activity` | Search logins, failed logins, sudo commands and file changes, with a time range |
| `search_ssh_keys` / `search_sudo_rules` | Search SSH keys and sudo privileges with identity mapping |
| `search_certificates` | Certificates discovered on your servers (expiring, weak, self-signed) |
| `search_pki` | Certificates ManageLM issues and renews |
| `search_monitors` / `search_backups` | Monitors with their current status, backups with their last run |
| `search_schedules` | Scheduled tasks with their last run and what it reported |
| `search_credentials` / `search_keystore` | Rotating credentials and keystore keys (metadata only) |
| `list_connectors` / `search_cloud` / `get_cloud_info` | Hosting connectors and the resources they discover |
| `cloud_action` | Start, stop, reboot or snapshot a VM (disruptive actions ask for confirmation) |
| `run_security_audit` / `run_inventory_scan` | Run a security audit or an inventory scan on one or more servers |
| `run_access_scan` / `run_certificate_scan` / `run_activity_scan` | Run an access, certificate or activity scan |
| `get_task_status` / `get_task_history` | Status of a task, recent tasks on a server |
| `get_task_changes` / `revert_task` | View file diffs from a task, undo them |
| `answer_task` | Answer a question from an interactive task |

## Portal

<p align="center">
  <img src="assets/portal.png" alt="ManageLM Portal — server management, health monitoring, command history" width="700">
</p>

## Self-Hosted

Replace `app.managelm.com` with your own portal URL in the MCP configuration. With the plugin installed, set `MANAGELM_MCP_URL` instead (e.g. `export MANAGELM_MCP_URL=https://portal.example.com/mcp`) before starting Claude Code: the plugin uses it in place of the SaaS endpoint. See the [self-hosted guide](https://www.managelm.com/doc/) for Docker deployment.

## Requirements

- **Claude Code 2.0.12+** for the plugin, or any MCP client with remote HTTP servers (Claude Desktop, claude.ai)
- **ManageLM account** — [sign up free](https://app.managelm.com/register) (up to 10 agents)
- **ManageLM Agent** — installed on each server you want to manage

## Other Integrations

ManageLM works with your favorite tools:

- [VS Code Extension](https://github.com/managelm/vscode-extension) — `@managelm` in Copilot Chat
- [ChatGPT Plugin](https://github.com/managelm/openai-gpt) — manage servers from ChatGPT
- [n8n Plugin](https://github.com/managelm/n8n-plugin) — infrastructure automation workflows
- [Slack Plugin](https://github.com/managelm/slack-plugin) — notifications and commands in Slack
- [OpenClaw Plugin](https://github.com/managelm/openclaw-plugin) — OpenClaw integration

## Links

- [Website](https://www.managelm.com)
- [Full Documentation](https://www.managelm.com/plugins/claude.html)
- [Portal](https://app.managelm.com)

## License

[Apache 2.0](LICENSE)