Cutting Optimization via MCP and AI Agents
Sawvant ships an MCP server. Connect Claude or any MCP-compatible agent and optimize cuts through natural language.
The Model Context Protocol (MCP) lets AI models call external tools through typed schemas. The model discovers available tools, constructs the correct payload, and incorporates the response into its reasoning. No pasting API docs into chat windows.
Sawvant provides a hosted MCP server alongside the REST API. It exposes four tools: authenticate, optimize_cutting, check_job and get_result. The input schema of optimize_cutting covers sheets, parts, strategy, machine settings and cost tariffs. Any MCP-compatible client (Claude Desktop, a custom agent loop, or any framework supporting the protocol) can connect and use the optimizer as a native tool.
A typical interaction: you describe a project in plain language. "10 shelves at 800×300mm and 4 uprights at 600×400mm from standard 2440×1220 sheets, minimize sheets used." The model maps your requirements to the input schema, calls the Sawvant tool, and presents the structured layout response: sheets used, part placements and yield. Change a dimension or add parts, it calls the tool again with updated parameters.
Configuration is a single entry in your MCP config pointing at https://mcp.sawvant.com/mcp. The agent signs in through the authenticate tool, or you put an API key as a bearer token in the headers block. The model then has live access to the optimizer. Plan cuts on demand, compare the fast and thorough strategies side by side, or iterate on layouts within a conversation. Full tool schema and auth setup are in the MCP server docs.