Timbr canonical manufacturing manifest
timbr.canonical-schemaWhat shipsSchema 1.0 with structural and semantic validators. Every adapter projects from it.
application/json
Timbr's designs are meant to be made, which means reaching the machine or software you already run. This page is the honest map of that work: what we can produce a file for today, what is planned, and what is blocked and why.
Every export starts from a single canonical manufacturing manifest: parts with stable identifiers, materials, edge banding with an explicit dimension policy, and machining operations expressed as tool intent rather than machine codes. DXF, cut lists and G-code are projections of that manifest. A manufacturing decision is never recorded first in a derived file.
That matters because each format loses something different. If a bore depth lived only inside a DXF layer name, nothing could check it, and the day your importer wanted a different layer convention the depth would change with it.
A connector that reaches a machine is safety-critical. Publishing a long list of logos would be easy; publishing what has actually been proven is useful. So each target carries the evidence grade behind it, and no target can claim certified production on third-party documentation alone — that rule is enforced in code, not by convention.
ISO 6983-1:2009 states in its own scope that conformance does not guarantee interchangeability of control programs between machines. “It's just G-code” is not a plan.
Grade C is the most useful tier for discovering a format and the most dangerous for production. Grade D may open an investigation; it may never close one.
You can export these now. None of them is certified for production — check the archive's README before anything reaches a machine.
timbr.canonical-schemaWhat shipsSchema 1.0 with structural and semantic validators. Every adapter projects from it.
application/json
timbr.dxf-profileWhat shipsProfile-driven per-part DXF with configurable decimal encoding, horizontal-drilling representation flag and declared target-profile lineage. Not yet imported into any named vendor CAM.
.dxf
Real vendor profiles encode decimals as letters (HO9p5 = 9.5), so decimal encoding is a profile parameter, never a constant.
timbr.cutlist-csvWhat shipsMapping engine plus a Timbr default profile. Optimizer-specific mappings ship as unverified hypotheses until run through the real importer.
.csv
A generic CSV is not enough: delimiter, decimal format, field mapping, material codes and optimisation settings are all part of the adapter.
timbr.step-exportWhat shipsAnalytic B-Rep solids with bores, grooves and rabbets cut as real geometry. Every file is read back with the same kernel and checked for solid count, volume, bounding box and surviving cylindrical faces. Not yet imported into any named vendor CAM.
.step.stp
Promoted from fallback to a primary OEM path in spec v2: woodWOP 7.0+ imports STEP natively, 7.1 adds bore recognition, 8.0 adds 3D groove/pocket recognition. Analytic geometry is the whole point — a faceted export has no cylinders, so feature recognition finds nothing and every hole is redrawn by hand.
timbr.iges-exportWhat shipsThe same analytic solids as STEP, written as IGES and verified by reading back. For importers that read IGES more reliably than STEP.
.iges.igs
woodWOP has imported IGES since 7.0, alongside STEP.
timbr.stl-exportWhat shipsBinary STL at 0.1mm linear deflection. A MESH: it carries no cylinders, so no importer can recover a bore from it. For visualisation and collision checking, not for machining.
.stl
woodWOP added STL import in 7.1. Offered because it is asked for, labelled as a mesh because that is what it is.
timbr.dxf3d-exportWhat shipsThe same solids as STEP, tessellated to 3DFACE entities. A MESH: a bore is a ring of triangles, so nothing downstream can recover it as a hole. For targets that genuinely cannot read STEP.
.dxf
SCM documents 3D DXF alongside STEP/IGES/STL as Maestro source geometry. Offered for that case only — STEP carries features, this does not.
generic.iso-gcodeWhat shipsPost DSL and 3-axis drilling/routing output behind an explicit opt-in, with a mandatory operation report. No controller profile is certified; every file carries an uncertified banner.
.nc.tap.prg.txt.cnc.anc.plt
'ISO' and 'Fanuc-compatible' are families, not a complete dialect. M-codes, canned cycles and macros are machine-builder specific.
tpa.tcnWhat shipsHoles on all six faces, programmed by diameter so TpaCAD resolves the tool from its own technology database. Routed features are reported, never approximated. Nothing has been opened in TpaCAD.
Coversfelder, format4, busellato, vitap, biesse-brema, biesse-vector
.tcn.f4g
Highest machine-coverage-per-engineering-month in the portfolio: one adapter reaches five OEM badges. This is why it is P0 despite shipping nothing yet. Spec v2 correction C2: Felder does NOT run generic ISO G-code. U4 is now largely answered for DRILLING: the vendor documents the hole working in full. A complete opcode dictionary for routing is still outstanding.
nesting.optimizersWhat shipsCut-list CSV with material, thickness, grain, quantity, label and part id. Column mappings for named optimizers are unverified hypotheses.
Do not encode drill or pocket operations in nesting contour files unless the optimizer explicitly supports them. Keep the machining packet linked by part id.
buildingsmart.ifc4What shipsIFC4 via IfcOpenShell, verified by reading the file back. Carries placement, size, material and the Timbr part id — NOT machining.
.ifc
For BIM and interior coordination. Never a CNC carrier — the spec is explicit that IFC must not be the sole source of machining truth.
d2m.btlxWhat shipsParts and dimensions as BTLx 2.0, VALIDATED against design2machine's own published XSD — the only adapter whose output is checked against somebody else's formal definition rather than our own tests. Machining is not included: BTLx operations are timber joinery on reference sides RS1-RS6.
.btl.btlx
Correctly out of scope for panel furniture - it is a timber-construction format - but far cheaper to implement than v1 assumed: the XSD is public and BtlViewer is open source. Implementer trap: BTLx 'Height' means component THICKNESS.
grbl.desktopWhat shipsA GRBL controller profile in the post framework, uncertified and opt-in.
Desktop controllers lack industrial drilling blocks, aggregate semantics and machine-safe macros. Do NOT reuse industrial posts unchanged.
These are not forgotten. Each is waiting on a specific answer we will not guess at.
homag.tapioBlocked onU5 - partner programme commercial and technical requirements. This is a commercial conversation, not an engineering one.
Converts the HOMAG problem from 'reverse-engineer a proprietary format' into 'become a partner and ship an app'. SWOOD and Palette CAD already have comparable arrangements.
homag.woodwopBlocked onOnly the NATIVE .mpr/.mprx path is blocked — on U1 (.mprxe structure) and on legal review. The machine itself is reachable now: woodWOP 7.0+ imports STEP and IGES natively, 7.1 recognises bores, 8.0 recognises 3D grooves and pockets. The layer-coded DXF path remains blocked on U8 (which DXF Import profile the customer runs).
.mpr.mprx.mprxe.dxf.step
The published layer vocabulary (Z_xx, HOxx, NZxx, FZxx, S_xx, FK-) is grade C and is a hypothesis to verify against the ACTIVE profile, never a table to hard-code.
biesse.bsolidBlocked onOnly the native CIX->BatchRun->BPP chain is blocked. U2 is now answered and the answer is negative: BppLib carries no licence, so it cannot be used or ported. bSolid imports STEP as source geometry today.
.cix.bpp.cid
Direct-to-NC is possible (the control accepts BPP,DXF,CID,CIX) but a CIX loaded that way cannot be edited or debugged, so it is expert-only and disabled by default.
Known hazardsbSolid has historically mis-imported some BPP process options: Z and depth swapped on stepped vertical milling, negative step increments on horizontal stepping. Any regression suite must assert on stepped operations specifically.
scm.maestroBlocked onOnly the native .xxl/.xcs -> XConverter -> .pgm chain is blocked, on U3. Maestro imports STEP, IGES, STL and 3D DXF as source geometry today.
.xxl.xcs.pgm.pgmx
.xxl and .xcs load into Xilog/Maestro but do NOT by themselves make a machinable part. This is a hosted-conversion integration requiring a licensed install on the customer side. X/Y mirroring and Z sign convention are documented sources of silent part-flipping errors.
vectric.vcarveBlocked onU7 - whether .pp posts are redistributable or reference-only.
.pp.crv.dxf
Dual purpose. Vectric's .pp is the best publicly documented post DSL in existence, and its variance set (decimal separator, decimal places, per-variable format overrides, unit multiplier, file extension, rotary wrapping, dwell) is the minimum feature list Timbr's own post framework implements.
hexagon.alphacamBlocked onU6 - whether xAPI permits an unattended server-side integration under licence.
Buys breadth of machine coverage through one partner. The saving is coverage, not escaping per-machine certification.
holzher.nchopsBlocked onNative .hop is blocked: no public schema and no open-source implementation exists in any language. aCADemy imports DWG/DXF, and NC-HOPS takes 5-axis CAD/CAM through a post-processor interface.
.hop
NC-HOPS is a hub, not just an OEM tool - it drives Biesse, Busellato, HOMAG and SCM machines too. Renaming another format's output to .hop does not work; the structure genuinely differs.
Prioritised by how many machines each one reaches per month of engineering, not by how well known the brand is.
mozaik.cncPathwayExport part and BOM definitions into Mozaik; let its post library reach the machine
Mozaik's published file-type list (.cix, .mpr, .nc, .xcs, .xxl, .tcn, .tap, .prg, .txt, .cnc, .anc) independently cross-confirms the format universe in the spec.
hexagon.cabinet-visionPathwayPartner/API job import preserving Timbr part ids; DXF+CSV as a fallback that is not equivalent to a native job
Same vendor group as Alphacam - a combined partner approach may be more efficient than two conversations.
microvellum.toolboxPathwayIntegrate at project/part/manufacturing-data level and let Microvellum's Integrated Post Processor reach the machines
Geometry alone loses Microvellum's product intelligence, construction rules and machining tokens.
imos.ixPathwayPush configured order and manufacturing data through iX Connect or an agreed database/API interface
eficad.swoodPathwaySOLIDWORKS/SWOOD plugin or structured STEP plus metadata; SWOOD handles the certified machine posts
SWOOD Connect is a bookable tapio interface to HOMAG productionManager - the precedent for homag.tapio.
topsolid.woodPathwayCreate TopSolid documents and manufacturing features through the Automation API, then use certified WoodCam posts
autodesk.fusionPathwayAdd-in or importer creating components, setups and operations from the Timbr manifest; Fusion's own CPS posts reach the machine
Post processors are executable logic and safety-critical. Pin the post revision and machine configuration.
onshape.restPathwayREST API creating documents, parts and assemblies with Timbr part ids preserved as metadata
Best cloud-native CAD connector candidate. Document microversions and element ids must be pinned.
trimble.sketchupPathwayRuby extension importing assemblies as components with manufacturing attributes
Faces and edges alone do not encode thickness, machining depth, face side or edge banding. The manifest must travel alongside.
dassault.solidworksPathwayNative parts and assemblies through the COM API with custom properties for materials, edges and part ids
erp.odooPathwayCanonical order/BOM/work-centre events over REST and webhooks, with CSV/SFTP fallback
Never write directly to database tables.
homag.cutritePathwayMapped cut lists, board data, labels and part ids; consider reaching it via tapio intelliDivide rather than direct file integration
vdma.opcua-woodworkingPathwayConsume the companion specification's WwMachineType information model rather than a bespoke telemetry mapping
chinese.controller-tierPathwayPrimary research required before any work
Large global installed base on low-cost nesting routers, absent from v1 entirely. Evidence grade D: this tier needs primary research, and grade D may never close an investigation. RichAuto DSP posts commonly emit .plt.
Guessing a machining parameter is a safety incident, not a shortcut. These stay open until a vendor, a customer or a site survey answers them.
Peer products we learn from rather than connect to.
The single most valuable thing we can be told is the exact machine, controller and software a real shop runs. That reorders this page faster than any amount of desk research.
Tell us your stack →