Verlon AI MCP — Claude Code MCP server: features and setup
This MCP server lets coding agents inspect Verlon AI gates, logs, recommendations, experiments, and models. Its default tools are read-only.
claude mcp add --env VERLON_API_KEY=<VERLON_API_KEY> --env VERLON_BASE_URL=<VERLON_BASE_URL> --transport stdio verlon-ai-mcp -- npx -y @verlon-ai/mcpWhat Verlon AI MCP does
The server exposes six tools for viewing gate settings, request logs, Cortex recommendations, experiments, and available models with pricing and capability scores. A separate tool, switch_model, can change the model used by a gate, but it is registered only when the server starts with --enable-writes.
The MCP client starts the server through npx. You need a VERLON_API_KEY; VERLON_BASE_URL is available when connecting to a self-hosted Verlon instance. Writes are off by default, so the agent cannot switch models unless you opt in with the flag. That matters if you expect to change gate settings through chat.
What it helps with
- inspect Verlon AI gate settings
- find failed requests in the logs
- retrieve Cortex recommendations for a gate
- check experiment status and variants
- compare available models and their pricing
Example prompts for the agent
- List my Verlon AI gates and show their settings.
- Find recent failed requests in the Verlon logs.
- Compare the models available to my Verlon AI gates.
How to install Verlon AI MCP
Platform CLI
claude mcp add --env VERLON_API_KEY=<VERLON_API_KEY> --env VERLON_BASE_URL=<VERLON_BASE_URL> --transport stdio verlon-ai-mcp -- npx -y @verlon-ai/mcpReplace <…> with your own values
Platform CLI
codex mcp add verlon-ai-mcp --env VERLON_API_KEY=<VERLON_API_KEY> --env VERLON_BASE_URL=<VERLON_BASE_URL> -- npx -y @verlon-ai/mcpReplace <…> with your own values
Config file
{
"mcpServers": {
"verlon-ai-mcp": {
"command": "npx",
"args": [
"-y",
"@verlon-ai/mcp"
],
"env": {
"VERLON_API_KEY": "<VERLON_API_KEY>",
"VERLON_BASE_URL": "<VERLON_BASE_URL>"
}
}
}
}Replace <…> with your own values
Config file
{
"servers": {
"verlon-ai-mcp": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@verlon-ai/mcp"
],
"env": {
"VERLON_API_KEY": "<VERLON_API_KEY>",
"VERLON_BASE_URL": "<VERLON_BASE_URL>"
}
}
}
}Replace <…> with your own values
Platform CLI
gemini mcp add -e VERLON_API_KEY=<VERLON_API_KEY> -e VERLON_BASE_URL=<VERLON_BASE_URL> verlon-ai-mcp npx -y @verlon-ai/mcpReplace <…> with your own values
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about Verlon AI MCP
How do I install Verlon AI MCP in Claude Code?
Run claude mcp add mcp -e VERLON_API_KEY=<VERLON_API_KEY> -- npx -y @verlon-ai/mcp. Replace the placeholder with your Verlon API key.
What does Verlon AI MCP do?
It gives an MCP client tools to inspect Verlon AI gates, logs, recommendations, experiments, and models. Model switching is available when started with --enable-writes.
Can this server switch a gate's model?
Yes. Start it with --enable-writes to register switch_model. Without that flag, the available tools are read-only.
