Compare AI models
in seconds
Run any prompt against multiple models side-by-side. Get a shareable link. One command to install.
claude mcp add --transport http evvl https://evvl.ai/mcpSponsored by Knowatoa
Get started in 60 seconds
Paste one command
Copy the MCP add command into your AI agent — Claude Code, Codex, or Gemini CLI.
Agent connects your OpenRouter account
Opens a browser link, you click Authorize. That's it — no API keys to manage.
Run your first comparison
Just tell your agent what to compare — it calls the right tool automatically.
Get a shareable link
Results are viewable at a unique URL. Share with your team — links expire after 7 days.
What will you compare?
Just tell your agent what to compare.
Compare Claude, GPT-4o, and Gemini refactoring this function to use async/awaitFind which model writes the cleanest refactored code.
Run this system prompt through Llama 3.1 70B and Claude Sonnet and compare qualitySee if an open-source model matches a premium one at a fraction of the cost.
Compare 3 models extracting entities from this textGet a shareable link so your team can review results together.
Test this prompt against 5 models at onceStop testing models one at a time in the playground.
Compare GPT-4o and Claude writing a product launch emailShare the evvl link in Slack and let your team pick the winner.
Compare o4-mini and GPT-4o classifying these support ticketsDiscover that a smaller model might beat the flagship on your task.
Recent Comparisons
Common questions
Is Evvl free?
Yes. Evvl itself is free. OpenRouter bills you directly for model usage.
What is MCP?
Model Context Protocol lets AI agents use external tools. Evvl is an MCP server — your agent connects to it to run model comparisons on your behalf.
Which agents work with Evvl?
Claude Code, Codex, Gemini CLI, and any MCP-compatible client.
How many models can I compare?
200+ via OpenRouter — including GPT-4o, Claude, Gemini, Llama, Mistral, and more.
Where are results stored?
Share links are temporary — they expire after 7 days. No permanent data storage.
Who built Evvl?
Evvl was built by the team at Knowatoa.
Start comparing today
One command. No signup required.
claude mcp add --transport http evvl https://evvl.ai/mcp