Why we built MeshCrunch

3D content keeps getting heavier while the devices that render it — phones, headsets, browsers — stay bounded by memory and bandwidth. The gap between what artists author and what runtimes can afford is where projects stall.

MeshCrunch closes that gap. It takes polygon-heavy source assets and produces optimized meshes, LOD chains, and format conversions through a queue-backed pipeline. Postgres is the source of truth for every job, so a refresh, a crash, or a second device shows the same state. The queues are split by tier, so paid work runs on its own workers and is not held up behind free traffic.

We run on owned hardware with zero-egress object storage, which keeps the economics of moving large 3D files sane — and lets us keep a genuinely useful free tier.

The same pipeline is reachable from an AI client. Our MCP server lets an agent work with the assets in your account on a paid plan, metered by the plan you already have rather than by a separate allowance.

Have a pipeline you want to plug this into? Get in touch.