MCP Integration Quickstart

Add Sematryx to Claude Desktop, Cline, or any MCP-compatible agent. Your agent gains expert optimization — call one tool, get an optimal answer with full explanation.

Prerequisites

Sematryx API key — Get one free (100 solves/month, no credit card required)

Claude Desktop or Cline (or any MCP-compatible agent)

Claude Desktop

Step 1: Find your config file

macOS:~/Library/Application Support/Claude/claude_desktop_config.json

Windows:%APPDATA%\Claude\claude_desktop_config.json

Step 2: Add Sematryx to the config

claude_desktop_config.json

{
  "mcpServers": {
    "sematryx": {
      "command": "npx",
      "args": ["-y", "@sematryx/mcp-server"],
      "env": {
        "SEMATRYX_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}

Replace YOUR_API_KEY_HERE with your key from /dashboard/keys.

Step 3: Restart Claude Desktop

Quit and reopen Claude Desktop. You should see Sematryx tools in the tool picker (hammer icon).

sematryx_optimizesematryx_explainsematryx_comparesematryx_formulatesematryx_formulate_continuesematryx_formulate_statussematryx_formulate_complete

Step 4: Test it

Ask Claude:

“Use Sematryx to find the optimal learning rate and dropout for a neural net, where lr is between 0.00001 and 0.1, dropout between 0.0 and 0.5, minimizing validation loss.”

Cline (VS Code)

Step 1: Open Cline MCP settings

In VS Code: Ctrl+Shift+P (or Cmd+Shift+P on Mac) → Cline: Open MCP Settings

Step 2: Add Sematryx server

Cline MCP Config

{
  "sematryx": {
    "command": "npx",
    "args": ["-y", "@sematryx/mcp-server"],
    "env": {
      "SEMATRYX_API_KEY": "YOUR_API_KEY_HERE"
    },
    "disabled": false,
    "alwaysAllow": []
  }
}

Step 3: Test it

Ask Cline in a new task:

“Use the sematryx_optimize tool to find the optimal allocation of a $100k budget across 3 channels: email ($5k–$40k), paid search ($20k–$60k), social ($10k–$50k). Maximize expected conversions.”

Other Agents / Custom Integration

Sematryx runs a standard MCP server at mcp.sematryx.com. Any MCP-compatible agent can connect using HTTP/SSE transport.

HTTP/SSE transport config

{
  "mcpServers": {
    "sematryx": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/client-sse"
      ],
      "env": {
        "SSE_URL": "https://mcp.sematryx.com/sse",
        "AUTHORIZATION": "Bearer YOUR_API_KEY_HERE"
      }
    }
  }
}

Tool Response Format

sematryx_optimize

{
  "success": true,
  "optimal_value": -1.2345,
  "optimal_params": [0.00312, 0.18],
  "solver_used": "cma_es",
  "evaluations": 847,
  "dimensions": 2,
  "explanation": "CMA-ES selected for smooth 2D landscape. Found minimum
at lr=0.00312, dropout=0.18. Lower lr prevents gradient explosion;
moderate dropout avoids overfitting. Solution stable across 5 restarts."
}

Pay per solve — no subscription needed

The free tier includes 100 solves/month — enough to build and test. Beyond that, $0.01 per simple solve, $0.05 per complex solve (≥50 dims or ≥5K evals). Credits never expire.

See full pricing →

API Key Security

Store your API key in environment variables or a secret manager — never hardcode it in your prompt. For Claude Desktop, the env block in the config file is safe because it's only read server-side by the npx process, not sent to Claude.

Next Steps

Agent Demo Tutorial

Watch an agent use Sematryx to solve a portfolio optimization problem end-to-end

Full API Reference

Python SDK, REST endpoints, and advanced configuration