Kyma MCP server
{
"mcpServers": {
"kymaapi": {
"command": "npx",
"args": [
"-y",
"@kyma-api/mcp-server@latest"
],
"env": {
"KYMA_MCP_KEY": "<kyma_mcp_key>"
}
}
}
}{
"mcpServers": {
"kymaapi": {
"command": "npx",
"args": [
"-y",
"@kyma-api/mcp-server@latest"
],
"env": {
"KYMA_MCP_KEY": "<kyma_mcp_key>"
}
}
}
}{
"mcpServers": {
"kymaapi": {
"command": "npx",
"args": [
"-y",
"@kyma-api/mcp-server@latest"
],
"env": {
"KYMA_MCP_KEY": "<kyma_mcp_key>"
}
}
}
}{
"servers": {
"kymaapi": {
"command": "npx",
"args": [
"-y",
"@kyma-api/mcp-server@latest"
],
"env": {
"KYMA_MCP_KEY": "${input:kyma_mcp_key}"
}
}
},
"inputs": [
{
"id": "kyma_mcp_key",
"type": "promptString",
"description": "Kyma mcp key",
"password": true
}
]
}{
"mcpServers": {
"kymaapi": {
"command": "npx",
"args": [
"-y",
"@kyma-api/mcp-server@latest"
],
"env": {
"KYMA_MCP_KEY": "<kyma_mcp_key>"
}
}
}
}file .mcp.json (in your project root)file claude_desktop_config.jsonfile ~/.cursor/mcp.jsonfile .vscode/mcp.jsonfile ~/.codeium/windsurf/mcp_config.jsoninstall npx -y @kyma-api/mcp-serverversion 1.0.3
What it does
Kyma API MCP Server exposes a live catalog of 120+ LLM models with real-time pricing, uptime measurements, and spending controls. It allows MCP clients to query model availability and rankings, check measured uptime per model, and send chat messages through a unified endpoint with configurable per-connection spend caps.
- Query live model catalog, pricing, and 30-day uptime metrics to choose the most reliable model for a task.
- Send chat messages through any available model with automatic cost calculation and enforcement of a per-connection spending limit.
- Monitor your spend and credits across model calls, with read-only tools that never incur charges.
Tools this server exposes
| Tool | Does |
|---|---|
| list_models | Returns the live catalog of available models with pricing and metadata. |
| get_model_uptime | Returns the measured 30-day uptime rate per model from real traffic and scheduled probes. |
| get_catalog | Returns pricing and model information. |
| get_rankings | Returns model rankings based on uptime and other metrics. |
| get_credits | Returns your current account credits. |
| get_spend | Returns your current spend against the per-connection cap. |
| get_requests | Returns your usage history and request details. |
| send_message | Sends a chat message through an available model, priced before execution and limited by the per-connection spend cap. |
| set_low_balance_alert | Sets an alert threshold for low balance. |
| get_topup_link | Returns a link to top up your account balance. |
Before you enable it
An MCP server runs with whatever access you give it. Start with a development account rather than production, and check which tools the server exposes before you let an agent call them unattended.