ServerManager
Agentless infrastructure management for servers and network devices, with a built-in AI copilot
In short
ServerManager is YGG's in-house agentless infrastructure management platform. Managed targets run no agent at all — every capability goes over an outbound SSH connection, and the control node can sit on a separate box, on a production server, or locally. Linux server management (CentOS / Ubuntu) is live and running in production; switch and router management (Cisco IOS / Juniper Junos / Huawei VRP) is under active development. A built-in AI copilot turns plain-language questions like "why is this device loaded?" into concrete diagnostic commands and explains the output.
What the agent takes off people's plates
Core capabilities
AI copilot
Turns plain-language questions into diagnostic commands and explains the output. Switchable across DeepSeek, OpenAI, Gemini and Anthropic — no single-vendor lock-in.
Agentless SSH executor
Nothing installed and nothing resident on managed hosts — everything runs over outbound, key-based SSH, with an xterm.js web terminal for shell access straight from the browser.
Network device management (in development)
Bring switches and routers under the same platform, covering vendor CLIs such as Cisco IOS, Juniper Junos and Huawei VRP: direct web-terminal access, scheduled configuration backup with version diffing, and batch rollout. Shares the credential vault and audit trail with servers, so network gear needs no separate tool.
In development
Cross-host patch auditing
Scans security-patch status across many hosts and applies fixes in one pass — the main gap this fills versus general-purpose panel tools.
Encrypted credentials and audit trail
SSH credentials are encrypted at rest under a master key instead of scattered across laptops; every connection and command is written to an audit trail attributable to a person.
Certificate issuance and renewal
Automatic Let's Encrypt issuance and renewal; the issuance dialog can pull domains from configured sites and expand them into the SAN list.
Container and site management
Docker/Compose orchestration, Nginx site configuration, and installation of common packages — including service-level install and start/stop for MySQL/MariaDB and Redis.
Side by side
How ServerManager differs from the two common approaches
| Dimension | ServerManager | General-purpose panel | Manual SSH + scripts |
|---|---|---|---|
| Footprint on managed host | Nothing installed, nothing resident | Panel or agent installed on each host | Nothing installed |
| Where the control plane lives | Anywhere — separate box, production server, or local | Usually on the managed host itself | The operator's own laptop |
| Fleet-wide batch and patch auditing | Built-in batch commands and cross-host patch auditing | Mostly single-host oriented; limited fleet coverage | Requires writing and maintaining your own scripts |
| Credentials and audit | Encrypted at rest under a master key, with a full action audit trail | Stored locally by the panel; audit granularity varies | Keys scattered across laptops, no audit trail |
| AI-assisted diagnosis | Built-in copilot, switchable across four providers | Generally not offered | None |
| Network devices (switches / routers) | Cisco / Juniper / Huawei support in development, sharing the credential vault and audit trail with servers | Not covered — a separate NMS is required | Log in device by device; configurations are not version-controlled |
How the agent and human supervision collaborate
System architecture (conceptual)
human
Operator (browser)
agent
Control node: Caddy + Next.js + FastAPI
agent
AI copilot (multi-provider)
agent
SSH executor (outbound, key-based)
output
Managed hosts: CentOS / Ubuntu (zero footprint)
Flow
- Operator (browser)→Control node: Caddy + Next.js + FastAPIHTTPS / WebSocket
- Control node: Caddy + Next.js + FastAPI→AI copilot (multi-provider)Plain-language question
- AI copilot (multi-provider)→SSH executor (outbound, key-based)Generated diagnostic command
- Control node: Caddy + Next.js + FastAPI→SSH executor (outbound, key-based)Direct action
- SSH executor (outbound, key-based)→Managed hosts: CentOS / Ubuntu (zero footprint)Outbound SSH
Verifiable facts
Numbers and sources
687 backend unit tests pass, 1 skipped (an integration test that must be explicitly enabled)
All 10 key APIs on the live instance return 200 (servers / docker / nginx / jobs / audit / certs / groups / credentials / setup-status)
Front-end and back-end i18n keys are exactly parallel between Chinese and English: 1186 = 1186
No agent is installed on managed hosts; every capability runs over outbound, key-based SSH
The AI copilot integrates DeepSeek, OpenAI, Gemini and Anthropic, switchable at runtime
Delivery and pricing
Pricing
One-time perpetual licence for self-hosted deployment; no cap on control nodes or managed targets
Priced by fleet size and delivery model — contact us for a quote.
Included
- · Every module — there is no community/enterprise split
- · Deployment and administrator training
- · Source-level private deployment; all data and credentials stay on your own machines
- · One year of version updates, including the network-device release when it ships
Provided by the client
- · Hardware and cloud resources — servers, network gear and the like
- · LLM API costs for the AI copilot — bring your own key, usage bills to your own account
- · Custom development and third-party system integration
Scan for the full solution document
Scan with WeChat to reach the solution lead and request the complete documentation and deployment checklist.
Where it fits
When this approach does not apply
- Network-device management (Cisco / Juniper / Huawei) is **under active development and not yet released**. What ships and runs in production today is Linux server management (CentOS / Ubuntu). If you need network gear covered right now, wait for the release or check the current status with us first.
- Database support is service-level only: install, start/stop and version detection for MySQL/MariaDB and Redis. **No schema operations, backup policies or slow-query analysis** — this is not a database administration tool.
- An honest note on maturity: the longest and heaviest production usage so far is YGG's own operations — there is **no history of large-scale multi-customer deployment yet**. Functionality and stability are covered by 687 automated backend tests plus live-instance inspection, but your environment will surface things ours has not. Pilot it outside your critical path first, then move it inward.
- Managed hosts carry no footprint, but the control node still needs a machine of its own (2 vCPU / 2 GB and up). If you run a single server and do not mind the panel sharing it, this architecture buys you little.
Further reading
Related applications
Ygg Lab Token — API Quota Gateway
Every model call, clear and in control — nine upstream provider types behind a single sk- key, with quota, logs and billing you own.
YggLab AI Agent Desktop
One desktop where 20+ CLI agents actually collaborate — assign work solo, form a team, or command them from your phone while work keeps moving 24/7.