Coin Identifier MCP Server

Vision-based coin and currency identification powered by Claude Vision. Send a photo of any coin, paper money, or bullion piece and get back structured identification data.

By Mowell Labs

What It Does

Tool	Description	API Cost
identify_coin	Photo → full coin/currency ID (country, year, denomination, mint, grade, value)	~$0.02
identify_item	Photo → auto-detect type (coin/currency/card/bullion) then ID	~$0.03
estimate_value	Coin attributes → web-searched market value from recent sales	~$0.01
estimate_melt_value	Metal + weight + spot price → melt value (no API call)	Free
estimate_silver_melt	Pre-1965 US coin denomination → silver melt value with grade premium	Free

Requirements

• Python 3.10+
• Anthropic API key (set as ANTHROPIC_API_KEY environment variable)

Install

$pip install coin-identifier-mcp

Or clone and install locally:

git clone https://github.com/mowell-labs/coin-identifier-mcp.git
cd coin-identifier-mcp
pip install -e .

Run

stdio (default — for Claude Desktop, Cursor, etc.)

export ANTHROPIC_API_KEY="sk-ant-..."
python -m server

SSE (for web clients, Apify, MCPize)

export ANTHROPIC_API_KEY="sk-ant-..."
fastmcp run server.py --transport sse --port 8001

Claude Desktop config

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

{
  "mcpServers": {
    "coin-identifier": {
      "command": "python",
      "args": ["-m", "server"],
      "cwd": "/path/to/coin-identifier-mcp",
      "env": {
        "ANTHROPIC_API_KEY": "sk-ant-..."
      }
    }
  }
}

Tools

identify_coin

Identify a coin or paper currency from a photograph.

Input:

• image_base64 (string, required) — Base64-encoded image data
• mime_type (string, default "image/jpeg") — Image MIME type

Output:

{
  "success": true,
  "coin": {
    "country": "United States",
    "denomination": "Quarter",
    "year": 1964,
    "mint_mark": "D",
    "composition": "90% silver, 10% copper",
    "variety": "",
    "errors": "",
    "grade_estimate": "VF",
    "condition_description": "Moderate wear on high points, all details visible",
    "name": "Washington Quarter",
    "description": "1964-D Washington Quarter, 90% silver",
    "estimated_value": 8.50,
    "is_graded": false,
    "grading_service": "",
    "grade_number": "",
    "certification_number": "",
    "sub_type": "",
    "friedberg_number": ""
  }
}

identify_item

Auto-detect whether image is coin, card, or bullion, then identify.

Input: Same as identify_coin.

Output: Same as identify_coin with added item_type field.

estimate_value

Look up current market value via web search.

Input:

• name (string, required) — e.g. "Morgan Silver Dollar"
• year (int, optional)
• denomination (string, optional)
• mint_mark (string, optional)
• grade_estimate (string, optional)
• variety (string, optional)
• grading_service (string, optional) — NGC, PCGS, etc.
• grade_number (string, optional) — MS-65, AU-58, etc.
• certification_number (string, optional)
• sub_type (string, optional) — fractional_currency, etc.
• friedberg_number (string, optional)
• country (string, optional)

Output:

{
  "success": true,
  "estimated_value": 42.50,
  "analysis": "Based on recent eBay sold listings...",
  "source": "web_search"
}

estimate_melt_value

Pure calculation — no API calls.

Input:

• metal (string) — gold, silver, platinum, etc.
• weight_oz (float) — Weight in troy ounces
• spot_price_per_oz (float) — Current spot price
• purity (float, default 0.999)

Output:

{
  "metal": "gold",
  "weight_oz": 1.0,
  "purity": 0.999,
  "spot_price_per_oz": 2350.00,
  "melt_value_usd": 2347.65
}

estimate_silver_melt

Pre-1965 US silver coin melt value with grade premium.

Input:

• denomination (string) — dime, quarter, half dollar, dollar
• year (int) — Must be pre-1965
• silver_spot_per_oz (float) — Current silver spot price
• grade_estimate (string, optional) — Grade for premium multiplier

Output:

{
  "denomination": "quarter",
  "year": 1958,
  "silver_content_oz": 0.1808,
  "silver_spot": 30.0,
  "melt_value_usd": 5.42,
  "grade_estimate": "VF",
  "grade_multiplier": 2.0,
  "estimated_value_usd": 10.85,
  "note": "Melt value is floor. Numismatic premium varies by date, mint, condition."
}

Environment Variables

Variable	Required	Default	Description
ANTHROPIC_API_KEY	Yes	—	Your Anthropic API key
COIN_ID_VISION_MODEL	No	claude-sonnet-4-20250514	Model for vision identification
COIN_ID_SEARCH_MODEL	No	claude-haiku-4-5-20251001	Model for value web search

Test Commands

# Test with fastmcp CLI
export ANTHROPIC_API_KEY="sk-ant-..."

# List tools
fastmcp dev server.py --list-tools

# Test melt value (no API key needed)
echo '{"metal": "silver", "weight_oz": 1.0, "spot_price_per_oz": 30.0, "purity": 0.999}' | \
  fastmcp run server.py --tool estimate_melt_value

# Test silver melt
echo '{"denomination": "quarter", "year": 1958, "silver_spot_per_oz": 30.0, "grade_estimate": "VF"}' | \
  fastmcp run server.py --tool estimate_silver_melt

# Test coin identification (requires API key + image)
python3 -c "
import base64, json
with open('test_coin.jpg', 'rb') as f:
    b64 = base64.b64encode(f.read()).decode()
print(json.dumps({'image_base64': b64, 'mime_type': 'image/jpeg'}))
" | fastmcp run server.py --tool identify_coin

Typical Agent Workflow

1. Agent receives coin photo from user
2. Call identify_item → gets structured ID + rough value
3. Call estimate_value with the returned attributes → gets market-verified price
4. (Optional) Call estimate_silver_melt for pre-1965 silver → gets melt floor
5. Agent presents complete identification + valuation to user

Cost Per Use

Operation	Model	Approximate Cost
Classify item type	Sonnet	~$0.005
Identify coin	Sonnet	~$0.02
Value lookup	Haiku + web search	~$0.01
Melt calculation	None	Free
Full pipeline	—	~$0.03-0.04

License

MIT

Platform Deployment

Apify

See apify/ directory (coming soon) for Actor wrapper.

MCPize

# Direct deploy via MCPize CLI
mcpize deploy server.py --name coin-identifier --transport sse