For one plant design (FINEP ThermoBiogasUp, Gen E), the flowsheets (P&ID drawn through an autocad-pid MCP on ezdxf, F02 Rev. D) sat in one branch and the 3D model (rev. 5, nozzle from-to list, hydraulic and gas notes) sat in another, each from a different session. This PR joins them, and the only conflict was .mcp.json, which now lists autocad-pid, build123d, freecad and blender. Two disciplines, two agent sessions, four tool servers, one git merge as the coordination step. Hydraulic tests 24/24 pass, but nothing checks that the P&ID and the 3D nozzles agree.