Coval MCP Server

What it does

Coval MCP Server bridges your AI assistant to Coval, an evaluation platform purpose-built for voice and text AI agents. With this server installed, Claude or Cursor can trigger evaluation runs against your agent fleet, manage test cases, inspect metrics, and iterate on your testing workflows — all through natural language, no dashboard-hopping required.

Tools

• list_agents — List all registered agents in your workspace
• get_agent — Retrieve details for a specific agent
• list_runs — Show recent evaluation runs with status
• get_run — Inspect a specific evaluation run
• create_run — Launch a new evaluation run against a test set
• list_test_sets — List available test sets
• get_test_set — Get full details of a test set
• list_test_cases — List all test cases within a set
• create_test_case — Add a new test case to a test set
• get_metrics — Retrieve scoring metrics for a completed run
• list_personas — List available evaluation personas

Installation

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "coval": {
      "command": "npx",
      "args": ["-y", "@covalai/mcp-server"],
      "env": { "COVAL_API_KEY": "your_api_key_here" }
    }
  }
}

Get your API key from app.coval.dev/settings.

Supported hosts

Confirmed for Claude Desktop and Cursor. Also supports remote HTTP connection via mcp-remote.