Verlon MCP server
{
"mcpServers": {
"verlon": {
"command": "npx",
"args": [
"-y",
"@verlon-ai/mcp@latest"
],
"env": {
"VERLON_API_KEY": "<verlon_api_key>",
"VERLON_BASE_URL": "https://..."
}
}
}
}{
"mcpServers": {
"verlon": {
"command": "npx",
"args": [
"-y",
"@verlon-ai/mcp@latest"
],
"env": {
"VERLON_API_KEY": "<verlon_api_key>",
"VERLON_BASE_URL": "https://..."
}
}
}
}{
"mcpServers": {
"verlon": {
"command": "npx",
"args": [
"-y",
"@verlon-ai/mcp@latest"
],
"env": {
"VERLON_API_KEY": "<verlon_api_key>",
"VERLON_BASE_URL": "https://..."
}
}
}
}{
"servers": {
"verlon": {
"command": "npx",
"args": [
"-y",
"@verlon-ai/mcp@latest"
],
"env": {
"VERLON_API_KEY": "${input:verlon_api_key}",
"VERLON_BASE_URL": "${input:verlon_base_url}"
}
}
},
"inputs": [
{
"id": "verlon_api_key",
"type": "promptString",
"description": "Verlon api key",
"password": true
},
{
"id": "verlon_base_url",
"type": "promptString",
"description": "Verlon base url",
"password": true
}
]
}{
"mcpServers": {
"verlon": {
"command": "npx",
"args": [
"-y",
"@verlon-ai/mcp@latest"
],
"env": {
"VERLON_API_KEY": "<verlon_api_key>",
"VERLON_BASE_URL": "https://..."
}
}
}
}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 @verlon-ai/mcpversion 0.4.1
What it does
MCP server for Verlon AI that exposes gates, logs, recommendations, and experiments as tools for coding agents. Lets developers inspect and manage Verlon AI infrastructure from Claude Code, Cursor, Cline, and other MCP-compatible clients.
- List and inspect AI gates configured in your Verlon account to understand routing and model selection.
- Review recent request logs filtered by gate, success status, or time range to monitor performance and costs.
- Switch which model a gate routes to without restarting, enabling model experimentation from your coding environment.
Tools this server exposes
| Tool | Does |
|---|---|
| list_gates | Returns all gates in the account with their id, name, description, model, task type, and creation date. |
| get_gate | Returns full configuration for a specific gate including model, fallback chain, spending limits, and orchestration. |
| list_logs | Returns recent request logs filtered by gate, time range, success status, with timestamp, cost, and latency. |
| get_recommendations | Returns Cortex intelligence report with optimization recommendations, drift detection, and themes for a gate. |
| list_experiments | Returns experiments (shadow and split tests) filtered by gate, status, or project, with variants and goal metrics. |
| list_models | Returns available chat models a gate can route to, with live pricing per 1M tokens and capability scores. |
| switch_model | Switches which model a gate routes to; takes effect on the next request. Requires --enable-writes flag. |
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.