Easy-CAD / bendcad (jnadeau207-collab)
An agent-native CAD kernel built so AI agents can design real parts alone and know when to believe the answer, with certified volumes, named refusals, and tolerance zones.
Summary rewritten from the vendor site because the catalog version did not match it.
What it does and how it connects
BendCAD is a new CAD kernel built for a different user than existing kernels: an AI agent that has to design real parts alone and know when to believe the answer. The README states that existing kernels were built for humans clicking buttons, and when they fail they fail quietly or with an error only an expert can read. BendCAD is written in Bend, is open source, and is presented as the CAD kernel AI agents can trust.
Its stated principles are that every number says what it is, proven, certified to a bound, estimated, or assumed, so nothing is silently wrong; failures explain themselves by naming what went wrong, where, why, and how to fix it, and leave the design untouched; each operation declares its intent and the kernel measures the result and rejects a mismatch; parameters carry tolerance zones and the kernel reports how answers move inside them; and work is parallel by construction.
The README lists shipped capabilities including a versioned B-rep archive, STEP read and write validated by independent readers, STL and 3MF mesh exports, assemblies with mates solved exactly, certified interference and clearance, semantic PMI and GD and T, sheet metal with certified flat patterns, and NURBS and Bezier patch solids with full validation. An agent or user sends edits to the bendcad CLI or its MCP server, and gets back a certified volume interval for every node, durable face names that survive parameter changes, and the sensitivity of the volume to each parameter.
Alternatives
Compare side by side- PartCAD
A package manager for parametric CAD parts versioned through git, rather than an agent-native CAD kernel with certified volumes.
- freecad-mcp (neka-nat)
The most widely used FreeCAD MCP server, with an addon plus server split and FEM analysis, rather than an agent-native CAD kernel.
- text-to-cad
An open source plugin that turns text prompts into CAD geometry with baked-in DFM rules, running locally inside the agent, rather than an agent-native CAD kernel.
- FrontierCAD (Normal)
A CAD benchmark from Normal Research's CAD Arena that scores how well AI agents recreate real parts across CAD tools, rather than a CAD kernel.
- RealCADBench
A benchmark for intent-to-program CAD modeling from real industrial design intents, scored on FreeCAD Python output, rather than a CAD kernel.
Sources
Not verified
- No GitHub releases exist, so latest release fields are null.
- No independent user voice found: the repo has no public issues.
- Pricing: none stated on the repo.
- The directory record describes founder-led research evaluating frontier AI models on CAD operations such as sweeps, lofts, and shells; the README as read describes the BendCAD kernel effort, and no model evaluation results were confirmed in the README.
Last verified November 6, 2026. Listings are informational. The evidence score measures public, verifiable evidence and is not an endorsement or a compatibility claim.