Frequently asked questions

Mesh simplification (or decimation) reduces the number of triangles in a 3D model while keeping its overall shape. MeshCrunch uses quadric-error metrics, which collapse edges in a way that preserves the silhouette. It is lossy by design: inspect the result before you ship it.

Accepted for upload: OBJ, STL, PLY, OFF, GLB, GLTF. A job can write back out as OBJ, STL, PLY, OFF, GLB — glTF is import-only, because a .gltf that references external buffers is not a single file. In-browser preview covers OBJ, STL, PLY, GLB, GLTF; there is no OFF loader for the viewer.

A level-of-detail (LOD) chain is a set of progressively simpler versions of the same model. Engines swap to a lighter LOD as an object moves farther from the camera, saving GPU time without a visible drop in quality. MeshCrunch builds the whole ladder from one request, every level decimated from the original rather than from the level above. Each generated level costs one processing credit, so a five-level chain has a base cost of five — multiplied, like every operation, by the source file's size band.

Up to 50 MB per file without an account or on a free one. Starter raises that to 250 MB and Pro to 1 GB.

Processing is metered two ways at once, and work needs room in **both**. **Processing credits** count operations, scaled by the size of the file they run on. The base cost is one credit for a simplify, one for a conversion, and one credit per level an LOD chain generates, so a five-level chain has a base of five. That base is then multiplied by the file's size band: Up to 25 MB is charged at the base rate, and above that a multiplier applies: 2× for 25–100 MB, 4× for 100–250 MB, 8× for 250–500 MB, 16× for 500 MB–1 GB. A 20 MB simplify is one credit; the same operation on a 180 MB model is four, and a five-level chain on it is twenty. **Processed data** counts the size of the mesh each of those passes reads. Without an account you get 3 credits per rolling 24 hours; a free account gets 10 per rolling 24 hours, Starter 500 per month and Pro 2500 per month. Alongside that, a free account may process 1 GB of mesh per calendar month and keep 250 MB of storage; Starter raises those to 25 GB and 5 GB, and Pro to 100 GB and 25 GB. On a paid plan both allowances reset with your billing period rather than on the first of the month. Browser-local preview and turntable export cost nothing at all — they never leave your browser. Every one of these is enforced by the API, not just shown in the interface.

Without an account, 24 hours — after that they are deleted automatically, and because the guest identity lives only in that browser nothing can reach them again. Signing in before then claims whatever you have uploaded onto your account automatically and re-stamps it with your plan's window: 7 days from upload on a free account, for the source and everything generated from it. You can also delete any asset yourself at any time, which removes the stored files immediately.

Drawn entirely by your own browser — no server renders anything and it costs no processing credits. The model itself is uploaded and stored as an asset, under the retention window for your plan or session. The same is true of the standalone viewer and GIF maker: the rendering is local, the file is not. Every upload goes directly from your browser or API client to private object storage rather than through our website, and we do not use uploaded models to train machine-learning models.

Converts validated OBJ, STL, GLB, glTF, PLY and OFF assets to OBJ, STL, PLY, OFF or GLB. The base cost is one processing credit, multiplied by the source file's size band. Formats that cannot carry materials in a single file say so before you convert.

Because an OBJ carrying materials is not one file: it is the .obj, a separate .mtl, and every texture image the material names. Handing back only the .obj would silently discard the rest, so the download is a ZIP of all of them plus a manifest. Conversions that genuinely produce one file — anything to GLB, or a model with no material data — download as that file.

Rendered and encoded by your own browser from GLB, glTF, OBJ, STL and PLY, with no account needed and no processing credits spent. The model itself is uploaded and stored as an asset like any other. You can also export MP4 where your browser supports it, and a PNG still with a transparent background — a GIF palette has a single transparent index and H.264 has no alpha channel, so those two are composited onto the backdrop you choose.

Rendered server-side to a private 512×512 image after upload, and served through short-lived authorized URLs. Shaded geometry only — textures are not rendered. They are deleted with the asset, do not count toward your stored-asset allowance, and cost no processing credits.

Yes, and not only from one. MeshCrunch runs a standard remote MCP server rather than a per-vendor integration, so any client that supports a remote MCP server over Streamable HTTP with OAuth — Claude Code, Codex CLI, Claude Desktop and Cursor among them — can work with the assets already in your account: list them, price an operation, run a simplify, an LOD chain or a conversion, and read the result. You connect it with your MeshCrunch sign-in and approve two permissions, and you can approve only the read one if you want an agent that cannot spend anything. Available on Starter and Pro plans. An agent is metered by the same plan you already have — the same processing credits, the same allowances — so connecting one raises no limit and buys no separate pool. Uploading is the one thing it cannot do. Assets are added in Studio, or through the REST API with a key on a paid plan.

Programmatic access is in beta on paid plans, through a REST API and the MeshCrunch MCP server. Both resolve to the same account, plan and quotas as the web app. REST callers authenticate with an API key created in Studio under Settings and can upload assets directly; MCP clients connect with your MeshCrunch login and work on assets already in the account. There is no SDK and there are no webhooks, so a caller polls for completion.

Read the REST API reference →

There is no service-level agreement, no guaranteed queue priority, and no choice of storage region. There is no single sign-on, no customer-visible audit log, no on-premise deployment, and no signed data-processing agreement. If you need any of those, tell us before you subscribe rather than after — we would rather hear it than have you find out after paying.