Skip to content

Server Management

Monitor, configure, and control every MCP server you deploy — from a fleet-wide overview to individual server tabs.


Infrastructure overview

The Servers page opens with a 4×2 metrics grid showing real-time performance across your entire fleet:

REQUESTS 24H
1,247
THROUGHPUT
P95 LATENCY
142
P95_MS
ERROR RATE
0%
ERROR_RATE
AVG RESPONSE
89
AVG_MS
SLOWEST TOOL
get_report
342MS
BUSIEST TOOL
list_users
891_CALLS
ACTIVE CLIENTS
3
CLIENTS_24H
LAST ACTIVITY
2m ago
LAST_REQ
MetricDescription
Requests 24hTotal MCP tool calls across all servers in the last 24 hours
P95 Latency95th percentile response time in milliseconds
Error RatePercentage of 4xx/5xx responses (green at 0%, amber at moderate, red above 5%)
Avg ResponseAverage response time across all tools
Slowest ToolTool with the highest average response time
Busiest ToolTool with the most calls in the period
Active ClientsCurrently connected SSE clients across all servers
Last ActivityTimestamp of the most recent request

Server list

Below the stats grid, every deployed server appears as a clickable row. Each row displays:

S
my-api-server
https://api.example.com · 3 SSE
ONLINE24 toolsDLPFinOps
S
skills-agent
Skills Server · 0 SSE
OFFLINE8 toolsLISTED

Server badges

BadgeMeaning
ONLINE / OFFLINEServer status with green/gray dot indicator
N toolsTotal number of MCP tools registered
DLPData Loss Prevention is active on this server
FinOpsFinOps Guard is active on this server
LISTEDServer is published on the Marketplace

Click any server row to open the server detail view.


Server detail tabs

Each server has five management tabs:


Frequently Asked Questions

How many MCP servers can I manage on Vinkius Cloud?

There is no hard limit on the number of MCP servers per account. Free plans allow up to 3 servers; paid plans scale to 10, 25, or unlimited servers depending on tier. Each server runs independently in its own V8 Isolate with separate metrics, tokens, and configuration.

What metrics are available in the fleet overview?

The fleet overview shows eight real-time metrics: Requests 24h (total tool calls), P95 Latency (milliseconds), Error Rate (percentage of 4xx/5xx), Avg Response (average latency), Slowest Tool (bottleneck identification), Busiest Tool (most-called tool), Active Clients (connected SSE sessions), and Last Activity (time since the most recent request).

Can I disable a server without deleting it?

Yes. Click the ONLINE/OFFLINE badge on any server to toggle its status. When disabled, the server stops accepting MCP requests and existing SSE connections are terminated, but all tokens, tools, configuration, and deployment history are preserved. Re-enable it at any time with a single click.

How do per-server settings work with global defaults?

Vinkius Cloud follows a global-to-per-server inheritance chain. Configure DLP patterns, FinOps Guard limits, and tool exposition once in Settings — every server inherits those defaults automatically. Then selectively override on individual servers when needed, without affecting the fleet-wide configuration.

Does Vinkius Cloud support multiple deployment methods?

Yes. You can deploy MCP servers via three methods: OpenAPI Import (paste a Swagger/OpenAPI URL), Manual API (define endpoints by hand for internal APIs without published specs), and Vurb.ts Framework (write structured MCP tools in TypeScript and deploy to the edge). All three methods produce servers that work identically with AI clients.

What AI clients are supported?

Any client that implements the MCP protocol. Built-in Quick Connect configuration is provided for Claude Desktop, Cursor, VS Code with Copilot, Windsurf, ChatGPT, Cline, Continue, Zed, and Antigravity. Paste the SSE URL and you are connected.