[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$f195n09vd0nr3p":3},{"code":4,"message":5,"data":6,"timestamp":55},10000,"success",{"full_name":7,"position":8,"description":9,"language":10,"stars":11,"forks":12,"stars_gained":13,"age_days":14,"hn_threads":15,"one_liner":16,"summary_md":17,"topic":18,"github_topics":21,"license":22,"homepage":23,"updated_at":24,"history":25,"related":26},"IPedrax\u002FBIMCraft",25,"BIMCraft: building information modeling; an open-source, clean-room take on the Revit workflow in pure Rust, built to be driven by agents","Rust",271,40,26,10,[],"A Rust BIM application that rebuilds Revit-style building modeling and can be driven by scripts or MCP agents.","BIMCraft is worth a look if you want a programmable BIM model you can generate, inspect and export without a commercial seat, but it is early and not a Revit replacement for production documentation yet.\n\n## The problem\n\nA designer or developer who wants a building model often starts in Revit: place levels, walls, doors, rooms, then produce plans, sections and sheets. The friction is that this workflow is tied to a proprietary desktop application, and automating a change typically means clicking through the UI or writing against an API that only exists inside that application. BIMCraft shifts the first step: a plain-text command script or an MCP client creates the same kinds of elements, and the result can be saved as a native project, an IFC4 file, a PNG plan or a PDF sheet set. The output is a model you can iterate on from code or an agent, not just a set of drawings.\n\n## What's different\n\nBIMCraft's distinguishing idea is that every action is a command. The desktop app exposes a Revit-like ribbon, Properties panel and Project Browser, while the CLI accepts one command per line such as `wall {\"points\": …}` or `door {\"at\": …}`. The README also documents an MCP server over stdio, including a mode where an agent controls the open desktop project through a local port, and changes appear in the window as undoable steps. That makes the same modeling operations available to a person, a script and an agent without a separate automation API.\n\nThe practical benefit is quick generation of schematic geometry and repeated variants: the `--highrise` sample builds a 20-storey, 9,521-element tower, and the sample house can be exported to a plan, a 3D view, a section and sheets from the command line. The trade-off is that this is a clean-room reimplementation in early development. The API-shape claim is about command coverage, not about matching Revit's full behavior or file fidelity, and the project's own status table is the place to check which element areas currently work. Import and export cover IFC2x3 and IFC4, but the README does not establish round-trip fidelity for arbitrary real-world models.\n\n## Who should try it\n\nA good fit is a developer or computational-design tinkerer who wants to generate building massing, rooms and walls from a script or an AI agent and then export IFC or drawings for inspection. It also suits someone evaluating agent-driven CAD workflows, because the MCP server and `--control` mode are documented paths rather than experiments implied by a badge. A poor fit is an architectural team that needs to produce permit or construction documents from an existing Revit model: the README labels the project early development, and the status table is candid about what works. The desktop app, CLI and MCP paths are separate entry points, so a headless user does not need the GUI, but the GUI is currently the way to control an already-open project interactively.\n\n## Requirements and limitations\n\nThe engine and CLI require Rust 1.90 or later, while the desktop app requires Rust 1.95 because of egui 0.36; there are no prebuilt binaries or package-manager instructions in the supplied README. You will build from source, and the default build includes the `rvt` feature, which contains code ported from rvt-rs and is Apache-2.0 only; a build with `--no-default-features` remains MIT OR Apache-2.0. That licensing detail matters if you redistribute a build or embed the reader, and the provenance document should be read before shipping it.\n\nOn the operational side, the desktop app performs a startup update check against GitHub unless disabled in settings or with `BIMCRAFT_NO_UPDATE_CHECK`; it also writes configuration beside your favorites. Autosave and recovery are documented, but the supplied material does not describe a full backup or versioning policy for project files, so plan your own. The `--control` mode bridges over a local port with a token written beside the favorites, which is convenient for an agent but another file path to understand. Performance figures are author-supplied in a baseline document, not independent measurements.\n\n## HotRepo's take\n\nHotRepo recorded BIMCraft at 271 total stars and #25 on the NEW-REPOSITORY board for Rust on 2026-10-26. That is a single snapshot from a 10-day-old repository, not a daily growth series, so persistence cannot be judged. The decision hinges less on the star count than on whether the documented element coverage matches the model you need; check the status table and try the sample with your own IFC file before committing a workflow.\n\n[Project source and documentation](https:\u002F\u002Fgithub.com\u002FIPedrax\u002FBIMCraft)",{"slug":19,"name":20},"hardware","Hardware & IoT",[],"Apache-2.0",null,"2026-10-26T00:46:34Z",[],[27,32,37,42,47,51],{"full_name":28,"one_liner":29,"language":30,"stars":31},"JellyFlow\u002FMcdonalds_in_RokidGlasses","A voice ordering assistant for Rokid Glasses that turns one spoken sentence into a verified McDonald's cart, with final submission left to a button press.","TypeScript",114,{"full_name":33,"one_liner":34,"language":35,"stars":36},"YosysHQ\u002Fyosys","An open framework that turns Verilog\u002FSystemVerilog RTL into gate-level netlists, for hardware designers and open-source ASIC\u002FFPGA flows.","C++",4840,{"full_name":38,"one_liner":39,"language":40,"stars":41},"apirrone\u002FOpen_Duck_Mini","Open hardware plans, 3D files and reinforcement-learning code for building and training a small bipedal duck robot at home.","Python",4475,{"full_name":43,"one_liner":44,"language":45,"stars":46},"NikeSong\u002FLeRobotWorkbench","A browser workbench that puts LeRobot robot-arm recording, training and evaluation behind clickable UI instead of CLI commands.","Java",105,{"full_name":48,"one_liner":49,"language":35,"stars":50},"Liu-zhongxian\u002Fzongzhu-bot","A round-screen ESP32-S3 desk voice terminal that listens on-device, runs PC commands locally, and forwards other requests to a Grok bot.",220,{"full_name":52,"one_liner":53,"language":35,"stars":54},"jamro\u002Ftiny-engineer","A 3D-printed desk robot that mirrors an AI coding agent's reading, thinking and writing through motion, faces and a bell.",865,1792984524]