# evalscopebench-mcp

A paid remote MCP endpoint that lets an agent run AI SDK benchmark gates, compare model scores and pull reports with receipts.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI
- Author: clauxel
- License: not stated · policy=link-only · description only; the file stays with the author
- Source: https://github.com/clauxel/evalscope-benchmark-mcp-mcp
- Updated: 2026-05-25
- HTML: https://affpapa.org/en/nevibecode/mcp/evalscopebench-mcp

## What it does

Four tools come with the endpoint: `run_benchmark_gate`, `compare_model_scores`, `read_benchmark_report` and `issue_benchmark_receipt`. An agent calls them to run a gate, line up scores from different models, or fetch a report. Everything returns as structured JSON, and each operation produces a receipt you can keep as workflow evidence.

Nothing runs on your machine. The endpoint sits at `https://evalscopebench.clauxel.com/mcp` and needs a client that speaks Streamable HTTP. A public server card is published at `https://evalscopebench.clauxel.com/.well-known/mcp/server-card.json`; the registry name is `com.clauxel.evalscopebench/evalscopebench-mcp`.

Access is paid. Production calls carry a bearer token issued from the product site, and an unauthenticated hit on `/mcp` just returns a JSON error. So the order matters: pick a plan, get the token, then wire the endpoint into your client. The repository holds documentation only — the service code stays private.

## Requirements
- Remote server: `https://evalscopebench.clauxel.com/mcp`

## Install

### Claude Code (native platform)

**Platform CLI: remote server**
```bash
claude mcp add --transport http evalscopebench-mcp https://evalscopebench.clauxel.com/mcp
```

### Codex

**Platform CLI: remote server**
```bash
codex mcp add evalscopebench-mcp --url https://evalscopebench.clauxel.com/mcp
```

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "evalscopebench-mcp": {
            "url": "https://evalscopebench.clauxel.com/mcp"
        }
    }
}
```

### GitHub Copilot

**Config file**
File `.vscode/mcp.json`:
```json
{
    "servers": {
        "evalscopebench-mcp": {
            "type": "http",
            "url": "https://evalscopebench.clauxel.com/mcp"
        }
    }
}
```

### Gemini CLI

**Platform CLI: remote server**
```bash
gemini mcp add -t http evalscopebench-mcp https://evalscopebench.clauxel.com/mcp
```

## When to use it
- run a benchmark gate before shipping an AI SDK build
- compare two models' scores on the same dataset
- pull a benchmark report as structured JSON
- keep a receipt for every benchmark operation
- expose benchmark checks to an agent over MCP

## Example prompts
- Run the benchmark gate on this build with `run_benchmark_gate`
- Compare model scores on this dataset using `compare_model_scores`
- Read the latest benchmark report and issue a receipt

## Similar
- [eval-engineering](https://affpapa.org/en/nevibecode/skills/eval-engineering.md) — skill · 5300 installs
- [gaia-submission](https://affpapa.org/en/nevibecode/skills/gaia-submission.md) — skill · 435 installs
- [cost-benchmark](https://affpapa.org/en/nevibecode/skills/cost-benchmark.md) — skill · 679 installs
- [llm-eval-harness](https://affpapa.org/en/nevibecode/skills/llm-eval-harness.md) — skill · 372 installs
- [deepeval](https://affpapa.org/en/nevibecode/skills/deepeval.md) — skill · 2800 installs
- [aiconfig-online-evals](https://affpapa.org/en/nevibecode/skills/aiconfig-online-evals.md) — skill · 2700 installs

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md