Agent Claude Code
Use to produce a human-readable summary of a CAD drawing. Given a local file path or a URL, calls the cadlens MCP tools to parse the drawing and reports its format, units, dimensions, layer breakdown, prominent text labels, and total entity count. Hand off to cad-layer-inspector if the user wants to drill into a…
Agent Claude Code
Use to inspect a single layer within an already-parsed CAD drawing. Given a jobid and layername, pulls all entities on that layer, characterizes its content (entity-type breakdown, bounding region, sample text), and proposes a semantic role (e.g., "wall lines", "dimensions", "title block").
Agent Claude Code
Use when the CADLens REST API returns an unexpected error (4xx/5xx) from a tool call and you need to reproduce, classify, and explain it against the contract in mcp-server-reference.md. Curls the API directly, decodes Zod validation envelopes, and maps status codes to root causes (revoked key, quota, DGN V8, oversized…
Agent Claude Code
Use when adding a new tool to the cadlens-mcp server. Scaffolds src/tools/ .ts, registers it in registry.ts, adds type defs to src/api/types.ts if a new REST endpoint is involved, and writes an integration test under tests/integration/.
Agent Claude Code
Use to validate that the cadlens-mcp server's stdio interface works end-to-end after edits to src/server.ts or src/tools/. Builds the project, pipes JSON-RPC tools/list and tools/call requests to the built binary, and verifies the responses match each tool's inputSchema.
MCP server Claude CodeCodexCursor
Model Context Protocol server that wraps the CADLens REST API. Runs locally from the @cadlens/mcp-server npm package. Needs 3 environment variables to run.