# Model Finder

Compare AI models against your own tests and find a lower-cost option among those that pass. Use it to assess candidates or prepare model routing.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI · ★ 3
- Author: Abhiindigloo
- License: MIT · policy=link-only · description only; the file stays with the author
- Source: https://github.com/Abhiindigloo/EvalHarness
- Updated: 2026-10-19
- HTML: https://affpapa.org/en/nevibecode/mcp/model-finder

## What it does

Model Finder suggests models for a task and budget, runs your examples against selected candidates, then reports pass rate, latency, and billed cost. It can also prepare routing that tries a cheaper model first and moves on when an answer fails quality checks. A separate skill estimates whether a local model fits your GPU.

The package includes four Agent Skills and a local MCP server built for Python 3.9+. The server itself makes no network requests and needs no API keys. Your own provider keys are needed when running evaluations or the generated router. Its output is advisory: it does not deploy anything or change the active model. To add it in Claude Code, run `claude mcp add model-finder -- uvx model-finder-mcp`.

## Requirements
- Needs: Python and uv (commands run via uvx)

## Install

### Claude Code (native platform)

**Platform CLI**
```bash
claude mcp add model-finder -- uvx model-finder-mcp
```
uvx runs the command named after the package; if the README names another one, use uvx --from model-finder-mcp <command>

### Codex

**Platform CLI**
```bash
codex mcp add model-finder -- uvx model-finder-mcp
```
uvx runs the command named after the package; if the README names another one, use uvx --from model-finder-mcp <command>

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "model-finder": {
            "command": "uvx",
            "args": [
                "model-finder-mcp"
            ]
        }
    }
}
```
uvx runs the command named after the package; if the README names another one, use uvx --from model-finder-mcp <command>

### GitHub Copilot

**Config file**
File `.vscode/mcp.json`:
```json
{
    "servers": {
        "model-finder": {
            "type": "stdio",
            "command": "uvx",
            "args": [
                "model-finder-mcp"
            ]
        }
    }
}
```
uvx runs the command named after the package; if the README names another one, use uvx --from model-finder-mcp <command>

### Gemini CLI

**Platform CLI**
```bash
gemini mcp add model-finder uvx model-finder-mcp
```
uvx runs the command named after the package; if the README names another one, use uvx --from model-finder-mcp <command>

## When to use it
- test candidate models with your own examples
- compare response cost and latency
- find a lower-cost model that passes your tests
- route failed requests to another model
- check whether a local model fits your GPU

## Example prompts
- Compare several models for this task using my examples, and show their cost and pass rate.
- Find a low-cost model for coding in my IDE and suggest how to test it with my prompts.
- Check whether this local model fits my GPU, including its KV cache.

## Similar
- [MCP Toolbox for Databases](https://affpapa.org/en/nevibecode/mcp/googleapis-mcp-toolbox-for-databases.md) — MCP server · ★ 16730
- [labelhead-artist-momentum](https://affpapa.org/en/nevibecode/mcp/labelhead-artist-momentum.md) — MCP server · ★ 107701
- [Agent Skills Search Server](https://affpapa.org/en/nevibecode/mcp/agent-skills-search-server.md) — MCP server · ★ 26673
- [Skyvern](https://affpapa.org/en/nevibecode/mcp/skyvern.md) — MCP server · ★ 23283
- [Nhost](https://affpapa.org/en/nevibecode/mcp/nhost.md) — MCP server · ★ 9369
- [PraisonAI](https://affpapa.org/en/nevibecode/mcp/praisonai.md) — MCP server · ★ 9298

---
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