@batchgrid/mcp

MCP (Model Context Protocol) server that exposes BatchGrid's batch inference capabilities to Claude Desktop, Claude Code, and other MCP-compatible AI tools.

Setup with Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "batchgrid": {
      "command": "npx",
      "args": ["-y", "@batchgrid/mcp"],
      "env": {
        "OPENAI_API_KEY": "sk-...",
        "ANTHROPIC_API_KEY": "sk-ant-...",
        "GEMINI_API_KEY": "AI..."
      }
    }
  }
}

Only include the API keys for providers you want to use.

Setup with Claude Code

claude mcp add batchgrid -- npx -y @batchgrid/mcp

Available Tools

| Tool | Description | |------|-------------| | analyze_data | Analyze a CSV/Excel file — column types, descriptions, task suggestions | | create_plan | Generate an execution plan from natural language intent | | estimate_cost | Calculate token count and cost estimate before running | | run_pipeline | Execute batch inference with checkpoint/resume support | | list_models | List available models and pricing across all providers | | batch_process | End-to-end workflow: analyze → plan → estimate → execute |

Example Usage

Once connected, you can ask Claude:

"Analyze the file customers.csv and classify each customer's industry"

Claude will use the BatchGrid tools to:

1. Analyze the file structure
2. Create an execution plan
3. Show you the cost estimate
4. Process all rows in parallel
5. Save the enriched output file

Environment Variables

| Variable | Provider | |----------|----------| | OPENAI_API_KEY | OpenAI (GPT-4o, GPT-4o-mini, etc.) | | ANTHROPIC_API_KEY | Anthropic (Claude Sonnet, Haiku, Opus) | | GEMINI_API_KEY | Google (Gemini Flash, Pro) | | OPENROUTER_API_KEY | OpenRouter (100+ models) |

License

MIT