Skip to main content
Glama
JonasKeil

model-version-mcp

by JonasKeil
README.md
# model-version-mcp

MCP server that provides up-to-date AI model names and pricing from [models.dev](https://models.dev).

Solves the problem of AI models hallucinating outdated model names by providing a live-fetched, filtered list of recently released models.

## Tool

### `get_ai_models`

Returns AI models released within the last 4 months with pricing data.

**Parameters:**
- `provider` (optional): Filter by provider name, e.g. `"anthropic"`, `"openai"`, `"google"`

**Returns:** JSON with model id, name, provider, release_date, cost ($/1M tokens), context window, and capabilities.

## Setup

### Claude Code (local)

```bash
git clone <repo>
cd model-version-mcp
npm install
npm run build
claude mcp add --transport stdio model-version-mcp node dist/index.js
```

### Claude Desktop

Add to `~/Library/Application Support/Claude/claude_desktop_config.json`:

```json
{
  "mcpServers": {
    "model-version-mcp": {
      "command": "node",
      "args": ["/absolute/path/to/model-version-mcp/dist/index.js"]
    }
  }
}
```

## Data source

[models.dev](https://models.dev) — community-maintained, no API key required. Cache refreshes every 60 minutes.

TDQS

A4/5.0

Scored across 1 tool

Disambiguation5/5

With only a single tool, there is no possibility of confusing it with another tool. The purpose of get_ai_models is clearly isolated and unambiguous.

Naming Consistency5/5

The sole tool name follows a consistent verb_noun pattern (get_ai_models), which is clear and descriptive. There are no mixed conventions since only one name exists.

Tool Count3/5

A single tool feels thin for an MCP server, even if the scope is narrowly focused on retrieving current AI models. While it can serve as a simple read-only utility, users might expect additional related operations given the server name.

Completeness3/5

The tool provides a core way to fetch recent AI models with pricing, but it lacks support for retrieving specific model details or older models outside the 4-month window. These gaps could force agents to work around the limitation or fail when needing historical model info.

Maintenance

ActivityInactive
ResponsivenessNo issues