Files
solidcraft/docs/mcp.md
T
+30 d4ec5e6811 Sync pinned main into render (#5869)
* Viewport: show reference geometry through bodies (Preferences, off by default) (#2697)

A new Graphics preference, "Show reference geometry through bodies", draws
inactive sketch contours, origin and construction planes and their axes over
shaded bodies. It is off by default, so the viewport looks as before: reference
geometry is depth-tested and a section leaves it whole (#1229); Create Sketch
and dialogs picking a plane still draw the planes through the model.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* ui-egui tests: offscreen GPU tests start their devices one at a time (#3989) (#3990)

* claim: #3989

* ui-egui tests: offscreen GPU tests start their devices one at a time

Every offscreen test started a wgpu instance of its own, and starting one
makes the Vulkan loader scan its drivers. With several test threads two
tests doing that at once crashed the test process with SIGSEGV inside the
loader (loader_icd_scan). offscreen_device now holds a process-wide lock
while it starts the instance, adapter and device; each test keeps its own
device and the rendering still runs side by side.

* DXF: write and read a Bézier at its own degree (#5269) (#5281)

* claim: #5269

* DXF: write and read a Bézier at its own degree (#5269)

The DXF writer and reader held a spline's degree to 7, the value MAX_BEZIER_DEGREE had before
#1642 raised it to 9. A G4 blend (one Bézier of degree 9) was written as a degree-7 SPLINE over
its first 8 control points, so a flat pattern's cut-out stopped 4.444 mm short in the file.

- seg_shape takes its limit from MAX_BEZIER_DEGREE: a Bézier is written whole, at its own degree.
- read_dxf reads SPLINE degrees 1 to MAX_BEZIER_DEGREE, the degrees a sketch spline takes.
- Tests: every Bézier degree 4 to 9, a degree-9 round trip, the reader's degrees, hostile SPLINE
  degrees and too few control points, the G4 cut-out exported and read back with its areas, and a
  degree-9 SPLINE with 2 to 10 points inserted into a sketch.

Fixes #5269

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* Renaming a plane keeps its projections, canvases and hidden state (#4353) (#4354)

* claim: #4353

* Renaming a plane keeps its projections, canvases and hidden state (#4353)

timeline.rename only renamed the feature in other features' references. A sketch's
projection of a construction plane, a canvas placed on it and the hidden list keep it
by name too, so they lost it: the sketch warned that the plane no longer exists, the
canvas left the view and a hidden plane showed again. Document::rename_feature_keys
renames it in sketch links and canvases, as rename_body_keys does for bodies; the
selection follows in the command, and the hidden list in Session::refresh_from, so a
hidden plane stays hidden through undo and redo too.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* Browser: add Rename and Delete to feature menus (#3839, #4343) (#4342)

* claim: #3839

* fix(ui): add Rename and Delete to browser feature menus

* Browser: a second click on a Document Settings entry closes its window (#4150) (#4152)

* claim: #4150

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Browser: a second click on a Document Settings entry closes its window (#4150)

Each entry under Document Settings only opened its window. A second click on
Units closed the menu on the press (the row is outside it) and opened it again
on the release; Precision left Preferences open; Parameters and Design
Properties came back as a fresh copy. Now a click on an entry whose window is
open closes it the way the window's own x does:
- Units: context_menu::toggle_for with the row as the opener, the fix #1522
  made for the workspace button's menu. A right-click still opens it at the
  pointer.
- Precision: prefs::close(app, true), the Cancel/x path that puts back what
  changed since the window opened (now pub(crate)).
- Parameters and Design Properties: the window is dropped, as its x does. A
  Parameters value typed but not applied yet is dropped with it, as today's
  second click already drops it.

Fixes #4150

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* Preserve expression-positioned sketch fillet corners (#4097) (#4106)

* claim: #4097

* Sketch fillet keeps a corner placed by an expression (#4097)

Filleting a corner whose position comes from parameter expressions deleted
the point and its expressions, so the sketch gained two degrees of freedom
and later parameter changes distorted the outline. keep_corner now also
keeps fixed or expression-positioned points as virtual sharp corners.

* Fix feature editor picks after upstream sketch resizing (#4083) (#4084)

* claim: #4083

* Feature editors refill named edges and faces where they are now (#4083)

After an upstream sketch change, Fillet, Chamfer, Shell and Draft evaluated
correctly, but their editors refilled the originally clicked points and
could then fail with "no edge at ...". Resolve the stored names through the
document's naming helpers first, keep the resolved index, and map the pick
into the displayed body's frame; the point lookup stays as the fallback.

* ui-egui tests: clear the texture deltas of the frames they run (#4039)

Nine tests dropped an egui frame's output with its texture delta still
unapplied (the expression completion editor's frames, the slot preview's
three, the browser's feature rows). epaint asserts against that in debug
builds, so they failed under the dev profile and passed only where debug
assertions are off. They clear the delta as the other tests do; one more
hover test that dropped its output the same way does too.

* Frames: profile library, members along frame sketch lines, and the Frame Member dialog (#4003) (#4004)

* claim: #4003

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Frames: the profile library (#4003)

The first slice of frames and weldments: the cross-sections structural members
will be made of, each SolidCraft's own parametric definition built in code
(no vendor or standard size tables are shipped).

- solidcraft_doc::frame: bar, round bar, rectangular and round tube, angle,
  channel, I-beam and a generic T-slot extrusion as exact line/arc regions,
  with defaults worked out from the profile's size, refusals that name the
  dimension that does not fit, anchors and placement, and exact section
  properties (area, centroid, Ixx/Iyy/Ixy by Green's theorem).
- frame.profile.list, frame.profile.properties (with mass per metre for a
  density) and frame.profile.insert (draws a profile into a sketch, locked by
  default, at a point, anchor and angle).
- Tests against hand-computed values (rectangles, circles, composite
  rectangles by the parallel-axis theorem, fillet spandrels), hostile input,
  extrusion volume, undo; catalog entries in every locale; docs/frames.md;
  ROADMAP M15.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Frames: drop the built-in T-slot profile (#4003)

Aluminium extrusions and other vendor sections come from the user's own
sketch (drawn or inserted from the vendor's DXF), so the library keeps the
plain steel shapes only.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Frames: members along frame sketch lines (#4003)

frame.member lays one body per straight line of a frame sketch (2D lines or
3D sketch lines), its cross-section square to the line from end to end:

- profile: a built-in profile, or profile_sketch, a sketch whose material
  (nested loops alternating, so two circles make a tube) is the section in
  its own x and y around its origin; profiles picks override it.
- anchor (origin, centre, centroid or a bounding-box corner or edge middle),
  angle (an angle expression, so a parameter), mirror.
- Orientation: the profile's y as near world up as the line allows, x
  square to both; a vertical member's x is world X (frame::member_axes).
- Members follow sketch and parameter edits, undo as one step, save and
  reopen (format: pending/frame-members.rs), and are named after the profile
  (Rect Tube 40x40x3, (2)…), numbering on past names already taken.

Tests against hand-computed volumes and bounding boxes (rectangle, 3D
tower lines, parameter edit, sketch tube profile and picks, anchor/angle/
mirror), refusals, undo/redo and reopen; walk fixtures and goldens; docs.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Frames: fixes from the adversarial review (#4003)

- Default member lines are what was drawn: projected and other linked curves
  and 3D wires (a face sketch's edges) and lines of no length are left out.
- A profile sketch's material starts from its drawn profiles, as
  solid.extrude's "all" does, so a section sketched on a face works; the
  even-odd test uses outlines and boxes worked out once, capped at 256.
- Member names are pinned to their lines (`names`, by line id; `body_name`
  is the base): deleting or adding a line, or deleting another feature, no
  longer renames other members, and frame.member stores every name it gives.
  body.rename renames a member on its line (also for later member features);
  a copy whose body_names were set still names its members by position.
- Built-in profile dimensions are expressions (ProfileSpec): they follow
  parameters, bare numbers take the design's units, and they are feature
  inputs. frame.profile.properties/insert read them the same way.
- Default root/toe radii shrink to the room a thick section leaves; radii
  given are checked by name; a radius must be 0 or at least 0.01 mm; error
  messages report the wall or flange value given, not twice it.
- A line within 1e-6 rad of vertical orients as vertical, so rounding never
  turns a post a quarter turn.
- A 2D frame sketch in a differently placed component is mapped into the
  member's frame, and member features recompute when placements change.
- Caps before lookups for a design's `lines`; a line listed twice is
  refused; a 3D wire that runs back on itself is not a line.

Regression tests for each, from the review's scenarios.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Frames: the Frame Member dialog on the ribbon (#4003)

- Frame Member (MODEL > Create) opens a dialog: the frame sketch's lines
  picked in the view (or All Lines), a built-in profile and its dimensions or
  a section picked in another sketch, anchor, angle, mirror and name, with a
  live preview. Edit Feature reopens it filled in.
- Lines picked in the view keep their sketch (a SKETCH_LINES input picks
  ExternalSketchCurve): line ids are each sketch's own, so two 3D sketches
  both have a `w2`, and the members went on the newer sketch's lines.
- timeline.redefine takes frame.member. Members keep their names through an
  edit unless it gives them a new body_name; frame.member takes `names`.
- A member whose section is a whole profile sketch keeps it when edited
  (the dialog shows the sketch instead of picking every region, which would
  fill a tube's hole).
- An original frame_member icon, open for review, on the ribbon, timeline
  and browser.
- The angle profile is listed as L-Angle, and web and flange as Web/Flange
  thickness, so they translate apart from the existing Angle, Web and Flange;
  the dialog's labels are in en, ja and zh. en-GB no longer repeats the frame
  command labels it spells the same.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* Rectangular and circular pattern: copies that miss the body are reported (#3042) (#3919)

* claim: #3042

* Warn when cut pattern copies remove no material

A rectangular or circular pattern whose copies miss the body succeeded silently and those copies did nothing. The pattern now reports how many of its cut copies removed no material, and keeps the copies that land.

* Delete in a rolled-back history warns about the features after the marker (#3471) (#3698)

* claim: #3471

* Delete in a rolled-back history warns about the features after the marker (#3471)

With the timeline rolled back, the checks a delete runs first (the
confirmation's "these features will fail", timeline.dependents and
selection.delete's would_fail) compared models evaluated only up to the
marker. Features after it, broken by the delete, were not listed, so their
error marks turned up only once the timeline moved to the end.

newly_broken_in_history evaluates both sides over the whole history when
either has a marker (on copies; the model and document are unchanged) and is
used by the delete's report and by timeline.dependents.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* macOS: Fusion navigation preset does not support Shift + two-finger orbit (#3446)

* Macos: keep touchpad/key modifiers true to fusion

* fix: remove obsolete orbit snap reset from trackpad navigation

Upstream removed ViewportState::orbit_snap. The new trackpad orbit path retained one reference, causing E0609 in the PR merge build.

Validation: all seven trackpad tests pass; cargo check -p solidcraft-ui-egui --all-targets passes on the exact failing CI merge 66a2791b with this fix applied; rustfmt and diff checks pass.

Co-Authored-By: GitHub Copilot <noreply@github.com>

* fix: retain Fusion trackpad help in refactored preferences

Preserve upstream's new preferences pages and apply the translated Fusion trackpad tooltip in prefs/navigation.rs.

Validation: cargo build -p solidcraft passed. Tests and full CI were not run, as requested.

Co-Authored-By: GitHub Copilot <noreply@github.com>

* style: format Fusion trackpad preferences tooltip

Apply rustfmt's required wrapping for the relocated tooltip. No behavior changes.

Validation: cargo fmt --all -- --check and git diff --check passed.

Co-Authored-By: GitHub Copilot <noreply@github.com>

---------

Co-authored-by: GitHub Copilot <noreply@github.com>

* 2-point circle takes a typed diameter (#3278) (#3335)

* claim: #3278

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* 2-point circle takes a typed diameter (#3278)

After its first point the 2-point circle tool showed no value box, so the
circle could only be sized with the cursor. It now shows a Diameter box,
like the centre circle: the typed value puts the second point that far
along the cursor's direction, and Enter draws the circle.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Phanoob <phanoob@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* Reorder design tabs by dragging them (#3187) (#3188)

* claim: #3187

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Reorder design tabs by dragging them (#3187)

A tab dragged along the strip follows the pointer while the others make
room, and lands where it is released; Escape, or releasing down the
window, leaves the order as it was. With the custom title bar a drag that
starts mostly up or down (or on the only tab) still moves the window.
The control channel's ui.documents gains a move action (from, to).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Phanoob <phanoob@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* Renaming a body keeps the evaluated timeline, off the window thread (#2467) (#2468)

* Renaming a body that later features change evaluates nothing (#2467)

Model::rename_body now keeps the results when later features change the
body (a pattern, a chamfer, a join). It still evaluates when a body of
that name comes and goes or isn't the same body throughout, when a later
body's name is worked out from it (X (2)), or when a message quotes it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Rename box: rename off the window thread, and locked bodies follow a rename (#2467)

The rename box commits body.rename through run_heavy, so a rename that
still evaluates later features (one names the body in its targets)
computes beside the window instead of freezing it. Locked bodies now
follow a body.rename from anywhere (the rename box, the control channel,
scripts), not only the rename box.

---------

Co-authored-by: bherbruck <brennen.herbruck@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* Toolbar: a constraint dragged out of the small box becomes a full button (#5254) (#5456)

* claim: #5254

* Toolbar: a constraint dragged out of the small box becomes a full button (#5254)

The Constrain panel's small box of constraint icons was a fixed list, so
dragging one onto the full-size buttons saved a new order but it was still
drawn in the box. The toolbar settings now remember, per tab, which box
commands were moved out; those are drawn as full buttons, the box closes
up, and dropping one back on a box icon returns it. Reset Toolbar restores
the default box, and settings saved before this load unchanged.

* Sketch offset keeps its source lines free to turn (#4464) (#4466)

* claim: #4464

* Sketch offset keeps its source lines free to turn (#4464)

An offset adds a Parallel between each source line and its offset line.
Since offsets became associative, the offset's points are held during a
solve, so each Parallel pinned its source line to the offset line's
current direction: a closed chain lost its turning freedom (8 degrees of
freedom became 6), and Horizontal or an angle edit on a source line was
refused as conflicting.

The solver and the constructive reading now leave out a Parallel that
ties an offset line to the source it follows while the association lives;
the association already keeps them parallel. After Break the stored
Parallel drives again.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* Flat Pattern view draws perforations (#4440) (#4441)

* claim: #4440

* Flat Pattern view: draw perforation holes (#4440)

The flat view drew the outline, cut-outs, pockets and bend lines but not
perforation fields, though the flat-pattern data and the DXF export list
their holes. The view now reads the hole centres from the sheet through
SheetBody::perforation_holes, caches them per model state and body, and
draws them as circles in the cut-out style. Holes outside the viewport or
smaller than one pixel are skipped, and the cache is capped at one million
holes.

* "Don't Save" looks like a button in the Save changes prompt (#4380) (#4400)

* claim: #4380

* "Don't Save" looks like a button in the Save changes prompt (#4380)

The prompt's third action used the kit's low-emphasis text style, so it
didn't read as clickable and people thought Cancel was the only way out.
Add an opt-in secondary_left style to the confirm footer (a framed
secondary button from the theme tokens) and use it for Don't Save.

* Save prompt test: clear each frame's texture delta once it is applied

egui 0.36 panics when a TexturesDelta with unapplied deltas is dropped;
the test kept only the last frame's output and dropped the earlier ones.

* 3D mouse test: clear the frame's texture changes, so it passes in debug builds (#4382)

each_frame_hands_the_view_its_motion_through_the_drivers_mapping dropped
the output of a frame it ran (`let _ = ctx.run_ui(...)`). epaint panics
in debug builds when a TexturesDelta is dropped with deltas left ("Dropped
TexturesDelta with 1 unapplied deltas"), so the test failed under a plain
`cargo test` on every platform while the ci profile passed. It now clears
them, as the other UI tests do since #60.

Co-authored-by: Tyler Chiriboga <18270301+Tyler-Chiriboga@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* Spanish (es-ES): catch up with en.json (#4292)

Co-authored-by: carlymx <24634598+carlymx@users.noreply.github.com>

* Esc ends the navigation bar's Pan, Orbit and Zoom modes (#4226) (#4233)

* claim: #4226

* Esc ends the navigation bar's Pan, Orbit and Zoom modes (#4226)

The nav bar toggled a mode on, but nothing cleared it on Esc, so a left drag
kept panning, orbiting or zooming. Esc now ends an active nav mode first;
the next Esc reaches the tool or dialog as before.

* Keep long Select Other labels inside the popup (#4209) (#4225)

* claim: #4209

* fix(ui): bound Select Other labels and expose full names (#4209)

* Deleting an inserted file removes its components, placements and joints (#4100) (#4113)

* claim: #4100

* Deleting an inserted file removes its components, placements and joints (#4100)

A STEP or mesh file inserted into a design makes a component tree. Deleting
its feature left that tree behind: the components, their occurrences, the
rigid joints that hold them and the bodies' component assignments. A later
body that took an imported body's name was then placed in the stale component.

Document::remove_component_tree removes a component subtree and everything
that refers to it: the features made inside it, its occurrences, the joints
on them with their motion links, joint origins, contact set members, motion
study keys and exploded moves, body component and offset entries, standard
part and link records, and configuration columns. Deleting an inserted
file's feature and component.delete both use it; component.delete now also
removes joints left pointing at a deleted occurrence. The body transfer
refusal moves with it, unchanged.

* Only an inserted file's feature takes its component tree when deleted

Mark the feature a STEP, IGES or mesh insert makes (`inserted`), and remove
the component tree on delete only for those and for a linked design's
feature. Deleting a copied component's independent base deletes just that
feature, as before. Older designs get the mark as they open (pending format
change), except copied components' independent bases.

* Rigid joint saved with the wrong occurrence when two occurrences share a component (#3857) (#3939)

* claim: #3857

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* Joint dialog: a body of a further instance picks that occurrence, not the component's first (#3857)

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>

* Cancel Sketch: leave a sketch without keeping its changes (#3838) (#3938)

* claim: #3838

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SBN24y2TcqRcm2zn8Gkj27

* Cancel Sketch: leave a sketch without keeping its changes (#3838)

sketch.cancel (Cancel Sketch, ribbon Exit group and the sketch's
right-click menu) puts the design back as it was when the open sketch
was created or opened for editing: a new sketch is removed with its
dimension parameters, an edited one returns to how it was. It is one
undo step, so Undo goes back into the sketch.

The session remembers the design and the undo history at the point a
sketch is entered. Cancel uses that only while the history still runs
through it (the same steps before, every step since taken in this
sketch); otherwise it refuses and points to Undo, so a stale entry can
never throw away later work.

The icon is the approved close glyph in the sketch accent (alias
sketch_cancel); ja and zh carry the new label.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01SBN24y2TcqRcm2zn8Gkj27

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* 3MF import: name a Bambu Studio project's bodies after its objects (#3858) (#3859)

* claim: #3858 (3MF import: names from a Bambu Studio project)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* 3MF import: name a Bambu Studio project's bodies after its objects (#3858)

Bambu Studio leaves the name off the root model's objects and keeps it in
Metadata/model_settings.config (<config><object id="N"><metadata key="name">),
so every object of a project came in as Mesh1, Mesh2, ... A build item whose
root object has no name of its own now takes the object's name from those
settings. A name in the model still comes first, and part names, other
settings and objects of other model parts are left alone.

The settings are read only when a body needs a name, within what the model
parts left of the 1 GB budget. Settings that can't be read (malformed, nested
too deeply, too large) never fail the import: the bodies keep the generic
names, with a warning saying why.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Tyler Chiriboga <18270301+Tyler-Chiriboga@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* Viewport: right-click keeps a multi-selection of faces and edges (#3733) (#3788)

* claim: #3733

* Viewport: right-click keeps a multi-selection of faces and edges (#3733)

---------

Co-authored-by: AhmetV <344941595+ahmetv10@users.noreply.github.com>

* Parameter import and image pickers offer their own file types (#3216) (#3314)

* claim: #3216

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Parameter import and canvas image pickers offer their file types (#3216)

Both opened the picker with no file type of their own, so it listed only
designs, CAD files and meshes (and All files). macOS lets you pick only
the types a filter names: CSV parameters, and PNG or JPEG images for a
canvas, were greyed out. They now ask for csv/json and png/jpg/jpeg,
which the picker lists first (#1801).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Phanoob <phanoob@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* Return to the originating tab after finishing a sketch (#3250) (#3264)

* claim: #3250

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* Return to the originating toolbar tab after finishing a sketch (#3250)

A sketch started from SHEET METAL or PLASTIC always ended on the SOLID
tab. Remember the tab that was active when the sketch began
(UiState::sketch_return_tab, not saved) and go back to it when the
sketch ends. Falls back to SOLID when that is not one of the normal
toolbar tabs. A sketch started from SOLID behaves as before.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* Test: no one-element loop in the return-tab test (clippy single_element_loop)

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>

* STEP import: let a cancelled import release the import slot (#3171) (#3189)

* claim: #3171

* Release cancelled file import slots (#3171)

* Dragging a part along its joint no longer highlights what's under the pointer (#5137) (#5140)

* claim: #5137

* Dragging a part along its joint no longer highlights what's under the pointer (#5137)

The viewport's hover pick was skipped during a box selection and camera
moves, but not while a jointed part was dragged, so faces and edges under
the pointer lit up on every frame of the drag. The pick now waits until
the joint drag ends.

* Browser: a body picked in the viewport is opened to and scrolled to (#3005) (#3081)

* claim: #3005

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Browser: a body picked in the viewport is shown in the Browser (#3005)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Merge main; viewport::select takes a toggle flag now

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Phanoob <phanoob@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* Line tool snaps a diameter through the circle centre (#4326) (#4327)

A line drawn across a circle only snapped its end onto the curve, so a
diameter could miss the centre at any angle. Near the centre the line now
locks onto the radial line: on the rim it stays on the circle and the centre
is kept on the line, and on the centre point the end lands there.

* Flatpak: find the host's fonts for CJK text and the font list (#5124) (#5127)

* claim: #5124

* Flatpak: find the host's fonts for CJK text and the font list (#5124)

Inside the Flatpak sandbox the host's fonts are mounted at /run/host/fonts
and /run/host/user-fonts, which no font search looked at. Japanese and
Chinese text fell back to boxes, and the font list missed the host's
families.

- Search both folders for the CJK and Chinese fallbacks and for the
  installed-font list.
- The CJK fallback reads SOLIDCRAFT_FONT_DIRS (a path list, as the font
  list already does), with CRAFT_FONTS_DIR kept as a fallback.

* Crash reports: keep component.activate in the journal (#5079) (#5121)

* claim: #5079

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Crash journal keeps component.activate (#5079)

Activating a component changes only the session's active component and
occurrence path, not the design, so the journal dropped the command and a
crash report's replay built later parts under the previously active
component. The journal now also keeps a command that changes either.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* Control channel and MCP on by default, with a setting to turn it off (#5561)

The app opens its localhost control channel on port 27878 (or the next free
one) unless Preferences > Assistant turns it off; --control and
SOLIDCRAFT_CONTROL_PORT still win. A current.port discovery file lets
solidcraft-cli mcp find the running app with no arguments.

Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>

* Tests: moving an occurrence recomputes only the cuts and joins it touches (#4965) (#4982)

* claim: #4965

* Joins and cuts follow only the placements they depend on (#4965)

Tests that turning a joint between parts cuts no hole again, and that a
tool cutting another part cuts again when that part moves relative to it,
while an unrelated part moving keeps the result.

Booleans and Combine now hash only the placements of the bodies they take
as inputs, relative to the frame that evaluates them (#3397), so these
pin that behaviour for joins and cuts with named targets across parts.

* fix(ui): preview shortcut chords and wait for complete modifier input (#1810)

Keep physical modifier events out of the ordinary-key slot, preview held keys immediately, and confirm after release and a one-second settling interval. Preserve legacy control-channel modifier behavior while allowing explicit physical Control/Command input. Add regression coverage, a UI scenario, localized guidance and documentation.

Fixes #1809.

Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>

* Move: scope box selection to the active component (#4459) (#4527)

* claim: #4459

* fix: scope Move box selection to active component (#4459)

* Delete removes a selected construction axis (#4692) (#4695)

* claim: #4692

* Delete removes a selected construction axis (#4692)

Delete on a construction axis skipped it with "the origin axes can't be
deleted". It now deletes the construction axis feature, like construction
planes; the origin axes X, Y and Z are still refused.

* Agent chat: end the agent's process group where `kill` is busybox' (#4587) (#4588)

* claim: #4587

* Agent chat: end the agent's process group with the system call (#4587)

The group was signalled with a `kill` program: `kill -s TERM -- -<pgid>`.
That works with procps' `kill` and with BSD's, but busybox' `kill` takes no
`--`: it prints its usage, exits 1 and signals nothing. The agent's helpers
(the CLI an adapter wraps, the MCP servers it starts) then stay alive after
the agent is ended.

The group is signalled with the system call instead, so no `kill` program and
no shell are needed and the behaviour is the same everywhere. `nix` (MIT) is
already in the tree through zip, and only on Unix: the browser build does not
take it.

The two tests that cover this now pass where they failed before:
`agent_chat::process::tests::dropping_the_process_ends_its_group` and
`agent_chat::process::tests::stopping_the_agent_leaves_no_helpers_running`.

* Extrude: picking a face, vertex or parallel plane off the profile switches the extent to To Object (#2855)

* Extrude: picking a face, vertex or parallel plane off the profile turns the extent to To Object

With profiles picked, a face off their plane (or curved), a vertex off it, or a
construction plane parallel to it is taken as the To Object target instead of
another profile: the extent switches without the Extent Type choice, and the
distance shows how far the target is. Coplanar faces still join the profiles.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Extrude: automatic To Object is a Preferences setting, off by default

Preferences > Design > "Extrude to a picked target" (`extrude_auto_to_object`, default
false) turns the automatic To Object on. Off, picks in the Extrude dialog only add
profiles and the distance is not synced to a target, as before this change.

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: bherbruck <brennen.herbruck@gmail.com>

* Regenerate THIRD-PARTY-LICENSES.txt and docs/features.md [skip ci]

* FreeCAD/Rhino import: decode caps with room for detailed parts (#1783) (#3209)

* claim: #1783

* FreeCAD/Rhino import: decode caps with room for detailed parts (#1783)

cadmpeg decoded FreeCAD and Rhino files under its service profile, whose
million collection items and 2.56e8 work units a detailed part (a sprocket,
a threaded rod, a part with a few hundred faces) runs past: the import was
refused with "resource limit on CollectionItems: BudgetExceeded". Allow
8 million items and 2e9 work units, half of cadmpeg's desktop profile; the
input-size cap stays SolidCraft's STEP limit and every other cap stays at
the service profile's.

decoding_is_capped checks the two raised caps (and that they stay below the
desktop profile's) and that every other cap is still the service one.

Fixes #1783

* A spherical shell revolves into a solid with a void, and cuts and joins with hollow tools work (#451) (#4182)

* Assembly: ground the first part; drag free joint chains (#2804) (#2805)

* assembly(feat): ground first parts and drag free joint chains

Ground the first top-level part when it is created and let face dragging reposition free occurrences or any member of a floating joint chain in one undo step. Preserve exact repeated-placement frames, joint values, modifier selection and Escape cancellation. Keep the accumulated fixture, viewport layout and documentation integration fixes from the existing PR.

Signed-off-by: Karl Von Richter<karl@embeint.com>

* assembly(fix): update movable fixtures for first-part grounding

Accept the documented grounded override in component creation, explicitly free movement fixtures, and keep linked clipboard pastes movable even in an empty document. Cover first and reused linked pastes, default insertion grounding, and single-step undo in the existing link tests.

Signed-off-by: Karl Von Richter<karl@embeint.com>

* assembly(test): keep the motion-cost crank free

Unground the first component in the document-level joint-face measurement fixture so its motion samples can rotate after first-part grounding. Preserve the rotation and shared-measurement assertions that caught the Linux CI failure.

Signed-off-by: Karl Von Richter<karl@embeint.com>

* assembly(test): free the frame placement fixture

Create the frame-member placement fixture ungrounded before moving it above its root sketch. Keep its world-space member assertions while respecting first-part grounding in the frame tests added on main.

Signed-off-by: Karl Von Richter<karl@embeint.com>

---------

Signed-off-by: Karl Von Richter<karl@embeint.com>

* UI: integrate theme radii, borders, palettes and viewport refresh (#5045) (#5046)

* claim: #4922

* UI: let themes own corner radii

* UI: use four-point panel and window radii in SolidCraft 95

* UI: align the 95 radius assertion with four-point panels

* claim: #4953

* UI: let themes own structural and control borders

* UI: single Blueprint outlines and dotted secondary rules

* UI: separate container outlines and add theme-owned 95 bevels

* claim: #5007

* UI: apply theme palettes to icons, workspace tabs and Gruvbox settings

* claim: #5045

* Viewport: invalidate cached geometry colours when the theme palette changes

* UI: preserve legacy theme logo colours with the borderless mark

* Section Analysis: OK with no plane picked turns the section off (#2410) (#2411)

While a section is on, Section Analysis opened with no plane picked now
says "No plane picked: OK turns the section off." and OK runs
inspect.section {"clear": true}. With no section on, OK with no plane
still asks for a plane.

Fixes #2410.

* Regenerate THIRD-PARTY-LICENSES.txt and docs/features.md [skip ci]

* Close Change Parameters with Escape (#3372) (#3399)

* claim: #3372

* Fix Escape cancellation in Change Parameters

Consume Escape before parameter editors can commit pending text on focus loss, while preserving autocomplete and shortcut capture priority. Add 11 UI regressions covering draft cancellation, preview cleanup, document and history preservation, and reopening saved values.

* Preserve construction references when axes and points are renamed (#4419) (#4449)

* claim: #4419

* fix: preserve renamed construction references (#4419)

* Browser: route body picks into open command dialogs (#2560) (#4371)

* claim: #2560

* fix(ui): route Browser body picks into active dialogs

* Fix midpoint picks in sketch dimensions (#4872)

Retain MID snaps across both picks and materialize reusable midpoint
constraints for point-to-line and point-to-point dimensions. Let midpoint
pairs choose horizontal or vertical spans with pointer placement before
entering a value.

Cover projected model edges, overlapping geometry, helper reuse,
undo/cancel/reopen, and existing angular and line-circle workflows.

* Sketch slots: provide typed dimension inputs for every variant (#2423) (#4318)

* claim: #2423

* fix(sketch): add typed dimensions to every slot tool (#2423)

* Version 0.21.0 on main (released from rc.2's code) (#5774)

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* docs(mcp): Windows setup for Claude Desktop (#3794)

* docs(mcp): Windows setup for Claude Desktop (paths, two servers, Store config file)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* docs(mcp): say the Store build can keep the config elsewhere, not that the other path fails

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* Regenerate THIRD-PARTY-LICENSES.txt and docs/features.md [skip ci]

* MCP: wait out long requests, honest batch outcomes, batch tool aliases (#2901)

Rebased onto main as one commit (was 120d26f7, df488f9a, db9d1073 plus
merges of main).

- A request the app reports as still running (#1618) is waited for on the
  same connection through `control.pending`; past the wait the error says it
  was not cancelled and will still apply (or only write its file), later calls
  are not sent behind it, and the new read-only `pending` tool returns its
  result once it lands.
- A failed batch says what really happened: `rolled_back` only when the
  design is back where the batch started (as #3288 / #3242 on main: the app's
  `rolled_back` is believed; for an older app, every completed step undone;
  only a rollback asked for is reported as one), plus `applied`, `not_run` and
  `failed.still_running`.
- A batch step may name a tool (set_parameter, list_commands,
  describe_command, execute, measure, inspect_design, undo, redo, export, save,
  open, new_design, crash_reports); screenshot, body_topology, batch and
  pending are refused before anything is sent.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* 3D mouse: driver choice, guided calibration, button mapping and a picture of the device (#1463) (#3825)

* Add a small test boat to steer around with the 3D mouse

A little hull in its own camera, drawn by the UI itself and not the CAD
viewport, so a 3D mouse can be tried out without opening a design. It
brings the model (generated by assets/models/make_test_boat.py) and the
tests for how the camera moves it.

Nothing uses it yet; Preferences picks it up with the rest of the 3D mouse
work.

* UI kit: keep a choice row from flipping with the scrollbar

A choice row picks segments or a dropdown by whether its labels fit. A page
scrollbar that comes and goes narrows the row by its own width, so the row
could flip back and forth between the two on a page that sits right at the
edge. It is now judged on the whole row and through a small helper,
fits_steadily, that only changes its answer once the miss is wider than the
scrollbar.

Also tests for that, for typed numbers in a slider's value box, and a
test hook to put up the issue-report input barrier by hand.

* 3D mouse: device model, guided calibration, buttons and a picture

The 3D mouse becomes a device we know something about instead of six raw
numbers.

The reader picks between spacenavd and the direct hidraw driver, falls back
from one to the other, and says so when it can't open the device. Guided
calibration asks whether the device is a spring cap or a trackball, adds a
touch step, and uses a settle filter so a ball let go of stops reading as
motion. Buttons get jobs (keys, held modifiers, commands), saved per
device, with defaults for the SpaceMouse Pro.

Preferences gets a picture of the device that moves as it does, next to the
test boat, and the same picture can sit in the 3D view at Left Viewport,
Right Viewport or Left Window, above the floating History card like the
legend and readouts.

These ride together because the crate's calibration types, the Preferences
page, the Linux reader and the button hook all use each other.

* 3D mouse: offer to calibrate an uncalibrated device

When a 3D mouse shows up that has not been calibrated, a card in the 3D
view offers to calibrate it, to swap drivers when the other one would work
better, to close, or to never ask again. It is made once per device, goes
away after 30 seconds in view, and gives way to a status or computing
toast instead of stacking on it.

The card is the shared toast, which now takes buttons: they group by
purpose, wrap inside the card at narrow widths, and do nothing while an
issue report holds CAD input. The picture of the device in the 3D view
keeps clear of the card.

* 3D mouse: translations and attribution

The 3D mouse strings (ui.3d_mouse_*, the calibration steps and the offer
card) in every shipped language, the Japanese catalog note, and the credit
for the test boat model in ATTRIBUTION.md.

This is last because the code before it only looks keys up: it builds and
runs without these, falling back to English until they land.

* 3D mouse: draw its preferences with the settings kit like the rest of Navigation

The 3D mouse section was a folded header with its own layout. It now
uses the same kit as the sections above it: a settings_section card,
settings_form rows with the shared label column, dividers between the
device, axes, picture and buttons groups, Text::Hint for the helper
lines, and settings_control_width for the dropdowns.

The section is open like every other one (the header still folds it),
so the device badge on the folded header is gone; the calibration row
already names the device. Every setting, string and key is unchanged.
Tests that unfolded the section no longer click it, and the hint lines
are looked up as hints instead of notes.

* 3D mouse: show each axis's live tilt in Preferences, like main's axis rows

Each view axis row is main's axis row again: a meter that follows the tilt of the motion driving it, with the dead zone marked, next to Invert, Speed and the "driven by" menu. The meter reads the calibrated travel when the device has a calibration and the raw reading over full scale when it does not, and asks for a frame whenever the reading changes. The helpers are back as they are on main, and the calibration state is a badge by the enable toggle.

* 3D mouse: line its preferences up with the rest of Navigation

- The two checkboxes sit side by side in the same two-column grid as Wheel and trackpad.
- Row labels stop repeating "3D mouse" and use sentence case: Calibration, Driver, Restore defaults, Reset view, Left viewport, Right viewport, Left window.
- Hints line up under the control column instead of the left edge.
- The six view axes are named Pan X, Pan Y, Zoom, Orbit X, Orbit Y and Roll so they fit the name column.
- Buttons are a divider and rows in a form like the other groups, without the bold sub-heading.
- The driver, dead zone, axes, picture and buttons sit behind a collapsed Advanced section, since most people never touch them.

* 3D mouse: tidy the Advanced test for clippy

* Timeline: suppressing a feature keeps later features' body names (#4302) (#4311)

* claim: #4302

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Timeline: suppressing a feature keeps later features' body names (#4302)

Bodies numbered by the counter (`BodyN`) take the next free number on every
recompute, so suppressing a feature that made one moved every later counted
body down one (`Body2` became `Body1`) and broke what named it, such as a
Move of `Body2`. timeline.suppress now stores the current names of later
features' counted bodies in their `body_names` first, as body.rename already
does for a renamed counted body (#1109). Engine only: no evaluation or
design-format change.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

* Timeline: a body made while a feature is suppressed keeps its name (#4302)

Suppressing a feature freed its counted name (`Body1`), so a body made
meanwhile took it and became `Body3` once the feature was back. Suppress
now also stores the suppressed feature's own counted names, and the
evaluator adds the names stored on earlier suppressed features to the
state's `reserved_body_names` (the names a rename freed, #4957), which the
`BodyN` counter passes by. They are taken from the design before each
feature is computed and are part of the state digest, so a cached result
is only reused when they match.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* cadmpeg 0.6.3: FreeCAD and Rhino import in the browser too (#4482)

cadmpeg 0.6.3 is pure Rust (ruzstd instead of the zstd C library, zip 9), so
solidcraft-io depends on it on every target and the wasm build reads .FCStd
and .3dm files like the desktop app. The browser's file pickers offer them.

0.6.3 replaces DecodeOptions::container_only with a decode depth; ask for
Geometry, what its STEP writer reads. zip 8 leaves the tree, and with it the
three cargo-deny advisory exceptions kept for it.

Co-authored-by: Claude Opus 5.5 (1M context) <noreply@anthropic.com>

* Pattern dialogs: Quantity fields get up/down step arrows (#4260) (#4264)

A new kit control, `ui_kit::count_field` / `count_row`, is an expression field for a
whole-number count with stacked step arrows at its right edge. An arrow steps the
evaluated count by one within the field's limits: a plain number is replaced, an
expression keeps its parameters and gains or adjusts a trailing `+ n`. The text
changes as typing changes it, so the live preview follows.

Used by the Quantity rows of the sketch Rectangular (1-500, both directions together
at most 500) and Circular (2-500) patterns, and the solid Rectangular, Circular and
Path patterns (1-1000).

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* Regenerate THIRD-PARTY-LICENSES.txt and docs/features.md [skip ci]

* Join: say so when the joined piece doesn't touch the body (#4434) (#4436)

* claim: #4434

* Join: say so when the joined piece doesn't touch the body (#4434)

A Join (extrude, revolve, sweep, loft, pipe, coil, primitives, thicken,
boundary fill, surface network) whose material meets none of its targets
still joins it to the first target as a separate piece. The result stays
as it was; the feature now warns: "the joined piece doesn't touch `Body1`,
so `Body1` is now in 2 separate pieces (New Body makes it a body of its
own)". History marks it and the command reply carries it.

The check compares pieces (outer shells, not voids) of the bodies the
feature changed or absorbed before and after it, so it holds for the
batched multi-profile path too, stays quiet for a join that bridges
pieces or closes a cavity, and leaves Combine (which joins the bodies it
is given) alone.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* Hole: snap to face and round-edge centres, drag placed holes, section in the dialog (#2171) (#2912)

Placing a Single hole put it exactly where the click landed, with nothing to line it up by and no
way to move it but taking it out and picking again; the dialog was a list of size fields.

- Preferences › Design › Create Hole has two settings, both off by default, so a new installation
  places holes as before:
  - Snap to face and edge centres (`hole_snaps`): the planar face under the cursor offers its
    middle and the centres of its round edges (a boss, a rounded corner, a hole already there) as
    marks; a pick near one, or on or beside such an edge, takes that centre and the mark is named.
    Sketch points on the face still come first; Alt or Ctrl held places it where clicked.
  - Move holes by clicking or dragging (`hole_moves`): each placed hole has a handle that drags it
    (snapping when snaps are on) and takes it out when clicked, and a pick elsewhere on its face
    moves it there. Edit Feature shows the handle too, so a hole can be moved later (#1047).
- The Hole dialog draws a section of the hole as its options make it, with each size in a field
  beside the dimension it sets; a hovered or focused field lights its dimension. The layout is a
  new kit element, `ui_kit::figure_rows` (a caller-painted figure beside captioned fields).

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

* Hole: place a Single hole by distances from reference edges (#37) (#2931)

A Single hole can take up to two straight edges of its face as references,
each with the distance of the hole's centre from that edge's line, measured
in the face's plane. `solid.hole` takes them as `references: [{edge, distance}]`;
the edges are found again by their persistent names, else their points, so
the hole follows them when the part changes, and each distance is a
parameter. A bad reference (no edge there, a curved edge, one square to the
face, parallel edges, a negative distance, a centre off the face) is refused.

In the Hole dialog, Reference 1 and 2 follow the face, each with its Distance
row; the canvas shows each distance as a dimension from its edge to the hole,
with a box just past the edge to type it in. Picking an edge measures its
distance (and its box takes the keyboard), typing one moves the hole, and
moving the hole measures them again. Edit Feature picks the edges again with
their distances.

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>

---------

Signed-off-by: Karl Von Richter<karl@embeint.com>
Co-authored-by: Frank <58932224+plexi32@users.noreply.github.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
Co-authored-by: raykholo <raykholo@users.noreply.github.com>
Co-authored-by: CamWight <91156642+CamWight@users.noreply.github.com>
Co-authored-by: SS-EN <84559909+SS-EN@users.noreply.github.com>
Co-authored-by: Colin Faulkingham <colin.faulkingham@gmail.com>
Co-authored-by: Karl Von Richter <karl@von-richter.net>
Co-authored-by: HiroKenEz <59478073+HiroKenEz@users.noreply.github.com>
Co-authored-by: Peppe Bergqvist <p@bergqvi.st>
Co-authored-by: GitHub Copilot <noreply@github.com>
Co-authored-by: phanoob <lesur.Stephane@wanadoo.fr>
Co-authored-by: Phanoob <phanoob@gmail.com>
Co-authored-by: Kai Evans <nemshilov@gmail.com>
Co-authored-by: bherbruck <brennen.herbruck@gmail.com>
Co-authored-by: Tyler Chiriboga <tylerchiriboga.apps@gmail.com>
Co-authored-by: Tyler Chiriboga <18270301+Tyler-Chiriboga@users.noreply.github.com>
Co-authored-by: CaRLymx <carlymx@gmail.com>
Co-authored-by: carlymx <24634598+carlymx@users.noreply.github.com>
Co-authored-by: Loohi <78334654+Loohi@users.noreply.github.com>
Co-authored-by: leValdemar19 <valdemarnord@hotmail.com>
Co-authored-by: AhmetV <mihendiz10@gmail.com>
Co-authored-by: AhmetV <344941595+ahmetv10@users.noreply.github.com>
Co-authored-by: rlartot <94693724+rlartot@users.noreply.github.com>
Co-authored-by: Christian Svensson <blue@cmd.nu>
Co-authored-by: Thodoris Tsagkogeorgas <thodoristsagko@gmail.com>
Co-authored-by: ahmetalpcelik <80778880+ahmetalpcelik@users.noreply.github.com>
Co-authored-by: ealitt <ealitt@gmail.com>
Co-authored-by: Homero Thompson <vdirienzo@gmail.com>
Co-authored-by: Diogo Soares Rodrigues <dsrodrigues20@gmail.com>
Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>
Co-authored-by: Artem Sushchev <Suschev.Artem+github@gmail.com>
Co-authored-by: Nathan Dane <nathandane.dev@gmail.com>
Co-authored-by: Kai Evans <hi@kaievans.co>
Co-authored-by: Victor Santos Oliveira <56417388+Vittuu@users.noreply.github.com>
Co-authored-by: cronos <crono2250@users.noreply.github.com>
Co-authored-by: rbousigues88 <rbousigues@gmail.com>
Co-authored-by: The Nomad <nomad@positron3d.com>
Co-authored-by: Jackson Oursland <jacob.oursland@gmail.com>
Co-authored-by: Jacques <59315319+Jacquesvdberg92@users.noreply.github.com>
2026-10-24 02:09:20 -04:00

32 KiB
Raw Blame History

SolidCraft and AI agents (MCP)

SolidCraft speaks the Model Context Protocol over stdio (JSON-RPC 2.0, one message per line; protocol revisions 2025-06-18, 2025-03-26 and 2024-11-05). An agent models exactly like a person does: every tool runs engine commands, the same ones the toolbar, palette, scripts and control channel use, and never opens a dialog.

Persistent headless and bridged sessions expose asynchronous product images through the ordinary execute tool: render.preflight, render.start, render.status, render.cancel and render.export. Keep the complete returned job ID; image bytes remain in the session until explicit PNG export. See render jobs.

For how the same command surface should work in the browser (workers, tiered storage, in-tab agents), see web-runtime.md. For sandboxed plugins and parametric timeline features that also use run(command, args), see plugins.md. Desktop MCP in this document remains the reference implementation today.

Start it

# Headless: an in-process design session, no window (screenshots are drawn on the GPU).
solidcraft-cli mcp

# Headless, starting from a design file or a command script.
solidcraft-cli mcp --in part.solidcraft

# Bridged to the running app, so you can watch the agent model. The app's control channel is
# on by default (see below), so no flag is needed on either side:
solidcraft &
solidcraft-cli mcp          # finds the running app; headless when none runs (--headless forces that)

# A fixed port, or a second app: name it.
solidcraft --control 7878 &
solidcraft-cli mcp --connect 127.0.0.1:7878

The app listens by default

The desktop app opens its control channel at start-up on 127.0.0.1:27878 (or the next free port when that one is taken), unless Preferences > Assistant > "Let AI agents connect (control channel / MCP)" is off; the setting takes effect at the next start. It is reachable from this computer only, and every connection must show the token made fresh for that launch, readable only by your user account (control-protocol.md). The app writes the port to control/current.port in the same folder as its token file; solidcraft-cli mcp with no arguments reads it. --control PORT and SOLIDCRAFT_CONTROL_PORT still choose the port and apply even when the setting is off. With the setting off and no flag, the app listens on nothing.

Bridged, the server finds the app's control token in the token file the app wrote for that port (control-protocol.md); --token T or SOLIDCRAFT_CONTROL_TOKEN give it explicitly when the app runs in a sandbox whose data folder the server can't see (prefer the variable: command-line arguments are visible to other users in the process list).

A bridged server starts even when no app listens on that port yet, so an MCP client that launches it before the app (or after the app was closed) keeps its tools. The server's instructions then say the app was not running, and every tool call answers with an error that starts no SolidCraft app is listening at HOST:PORT and says how to start it (solidcraft --control PORT). The first call after the app starts connects to it, reading that launch's token. When the app quits or crashes mid-session, the next call says the same; a call that was running when it went away says the app did not answer and may have quit or crashed, after running the call or not, so inspect the design before sending a change again. A batch that could not be sent says the app is not running, so it is safe to send again.

stdout carries only protocol messages; errors go to stderr. Bridged screenshots are written to a temporary file by the app and read back by the server, so both must run on the same machine.

Parameter checking

Commands with a typed parameter contract (solid.box, solid.extrude and engine.describe so far; describe_command shows a command's schemas) can be checked as they are called. --param-check off|warn|strict sets the level for a headless session, including a script or recipe given with --in; the default is off, which changes nothing. warn runs the call and adds param_warnings to its result, such as unknown key `operaton`; did you mean `operation`? (a starting script's warnings go to stderr). strict refuses the call, and the design is left as it was. Calls bridged to the app are not checked yet.

Which files the agent can reach

Tool calls read, write and delete files only inside the server's folders:

  • the folder of the open design (the --in design, or whatever the agent last opened or saved), but only when it lies strictly inside your home folder (or the temporary folder) and not in a hidden folder. Home itself, a folder above it (/, /home, /Users, a drive's root), hidden folders such as ~/.config and folders elsewhere don't count. Hidden files and folders inside the design's folder (.env, .git) are out of reach too;
  • SolidCraft's own recovery folder, so doc.recovery_list, doc.recover and doc.recovery_discard work out of the box (discarding deletes only real recovery entries);
  • the server's own scratch folder (a new private solidcraft-mcp-* folder in the temporary folder), where screenshot without a path goes;
  • every --allow-path DIR (repeat it for more). A folder named here is reached in full, hidden files included.

The working folder the server was started in does not count by itself, so ~/.ssh, ~/.aws, ~/.bashrc and other dotfiles are never in reach by default. stderr lists the folders at start-up and says when the --in design's folder was left out; --allow-path adds it back.

Paths are resolved before they are compared, so .. and symlinks that lead out are refused (saving through a link writes the file it points to, so the link's target is what counts), and a new file must go into a folder that exists (a missing folder inside an allowed one is reported as missing; it is not created and does not call for a wider --allow-path). On Windows a path must be on a local drive (C:\… or \\?\C:\…): UNC, device and NT paths in any spelling (\\server\share, //server/share, \\?\UNC\…, \??\…, \\.\…), drive-relative paths (C:file), a : in a file name (alternate data streams), reserved device names with any extension (NUL, CON, COM1.txt, …) and short 8.3 names (PROGRA~1) are refused before any file system call, and the resolved path is checked again, so a link or junction that leads to a share is refused too. This includes a sketch.text font given as a file path and the font file a design's text names when the design is opened; installed fonts named by family still work. The server spells its own folders (scratch, recovery) with long names, so they work when the temporary folder's path has a short name (…\NAME~1.XYZ\AppData\Local\Temp, as Windows often gives when the user folder's name is long). A refused path is a tool error saying why.

A hard link inside an allowed folder can't be told apart from an ordinary file, so a hard link to a file elsewhere is reachable through it; don't keep such links in a folder you allow.

solidcraft-cli mcp --allow-path ~/designs --allow-path ~/exports
solidcraft-cli mcp --allow-any-path   # no limit (the behaviour before this option existed)

Bridged, the server sends its folders with every request (allow_paths, see control-protocol.md) and the app adds its own recovery folder and the folder of the design open in it, under the same rules, so give absolute paths: a relative one resolves against the app's working folder, not the server's. The app's own file dialogs are never limited.

Tool annotations

Each tool says what it may do, for clients that ask before running tools (or auto-approve on these hints). list_commands, describe_command, inspect_design, measure, body_topology, skill_reference, crash_reports, pending and running are read-only. set_parameter, undo and redo change the design undoably. new_design replaces it (destructive). execute, batch, open, save, export and screenshot (whose path writes a file) can read, write, overwrite or delete files: they are not read-only, destructiveHint and openWorldHint are true.

Client configuration

Claude Code (project .mcp.json, or claude mcp add solidcraft -- solidcraft-cli mcp; a setup added under the previous name keeps working, see the compatibility notes):

{
  "mcpServers": {
    "solidcraft": { "command": "solidcraft-cli", "args": ["mcp"] }
  }
}

Claude Desktop (claude_desktop_config.json), bridged to a running app:

{
  "mcpServers": {
    "solidcraft": { "command": "/path/to/solidcraft-cli", "args": ["mcp", "--connect", "127.0.0.1:7878"] }
  }
}

On Windows the installer puts solidcraft-cli.exe next to solidcraft.exe in C:\Program Files\SolidCraft (the portable zip has both too). Give the full path, with each backslash doubled in JSON. This configuration declares two servers: solidcraft models headless (no window; steadier for timings), solidcraft-live drives the app you have open, so you can watch the agent model. --allow-path names a folder where the agent may save and export; it must already exist.

{
  "mcpServers": {
    "solidcraft": {
      "command": "C:\\Program Files\\SolidCraft\\solidcraft-cli.exe",
      "args": ["mcp", "--allow-path", "C:\\Users\\NAME\\SolidCraft-Agent"]
    },
    "solidcraft-live": {
      "command": "C:\\Program Files\\SolidCraft\\solidcraft-cli.exe",
      "args": ["mcp", "--connect", "127.0.0.1:7878", "--allow-path", "C:\\Users\\NAME\\SolidCraft-Agent"]
    }
  }
}

Start the app with its control channel for solidcraft-live, from PowerShell: & "C:\Program Files\SolidCraft\solidcraft.exe" --control 7878, or once for every launch: setx SOLIDCRAFT_CONTROL_PORT 7878.

Open claude_desktop_config.json with Claude Desktop's Settings > Developer > Edit Config rather than by its path: the Microsoft Store (MSIX) build of Claude Desktop can keep its data in its own package folder (%LOCALAPPDATA%\Packages\…\LocalCache\Roaming\Claude) instead of %APPDATA%\Claude, and Edit Config opens the file the app actually reads. Quit Claude Desktop completely (from the notification area) and start it again to load the servers.

Agent skills

skills/ holds four Agent Skills for agents that model through this server:

  • solidcraft-modeling covers building a new part.
  • solidcraft-editing covers changing an existing design.
  • solidcraft-checking adds checks for errors SolidCraft does not report, regression tests and porting, next to the other two.
  • solidcraft-mechanisms covers assemblies: joints placed by world coordinates, motion links, closed loops such as actuators, contact sets and motion studies, checked against analytic values.

They cover the workflow, the command shapes, how to verify with measure, and the known traps. With them an agent builds a part in a few batch calls and reads far less output. To install them for Claude Code, copy the folders into ~/.claude/skills/; skills/README.md has the details.

The skill_reference tool returns these documents from the copies built into SolidCraft, so an agent without file access can follow them: {"skill": "solidcraft-modeling"} returns its reference.md, the default file. Mechanisms also has reference.md, checking has reference/porting.md, and every skill has SKILL.md (editing uses modeling's reference). It never reads a path from disk, and an unknown skill or file is refused with the available choices. SolidCraft's own chat turns the agent's file tools off, so this is how its agent reads the skills there.

Tools

Tool Arguments What it does
list_commands filter? The command catalog: id, label, tab/panel, shortcut, parameter docs, enabled now. filter matches id, label, tab and panel.
describe_command id One command's parameter docs (params) and, for commands with a typed contract (solid.box, solid.extrude and engine.describe so far), the JSON Schemas (2020-12) of its parameters (input) and of a successful result (output); null for the others. The same as execute engine.describe {command}.
execute command, params?, verbose? (default true) Run any command with JSON parameters (list_commands shows them). Failures leave the design unchanged. A feature command replies with the feature (feature, name), every body (bodies) and the bodies it created, changed and removed (created_bodies, changed_bodies, removed_bodies). verbose: false uses the same compact reply as batch.
batch commands: [{command, params}], rollback? (default true), verbose? (default true), timings?, profile? (default false) Run steps in order, stop at the first error and undo the completed steps. The reply names the failing step (failed.index, failed.command, failed.error), every result before it, rolled_back (true only when the design really went back to where the batch started), undone, applied (the indices of the steps whose changes are in the design) and not_run (the steps never sent). A step may name a tool instead of a command, with that tool's arguments as params: set_parameter (runs parameters.change; value or expression), list_commands (engine.commands, filtered by filter as the tool filters it), describe_command (engine.describe), execute ({command, params}), measure, inspect_design (document.inspect: each sketch as its counts, not its full contents, and no joints), undo, redo, export, save, open, new_design and crash_reports. screenshot, body_topology (use model.edges and model.faces), batch, pending, running and skill_reference can't be steps: such a batch fails before anything is sent. verbose: false keeps the reply small in large designs: feature steps leave out the full body list (created_bodies, changed_bodies, removed_bodies stay) and sketch steps the ids they repeat outside created. Bridged to an app that has engine.batch, the whole batch is one control request (one frame of the app). Older apps, and apps showing the agent cursor, get the steps one by one. timings: true adds timings: [{index, command, ms}] for every step run, the failing one included, and total_ms. profile: true also adds each step's profile: {boolean_ms, tessellate_ms, other_ms}, its time in kernel booleans, in tessellation and the rest. Steps are timed where they run: in the app when bridged. An app without engine.batch is timed step by step, without profile.
inspect_design measure? (default true), sketches? (default true), joints? (default true) Design properties (metadata: project, author… and other tools' keys; set them with execute document.metadata.set), parameters, timeline (type, suppressed, rolled back, error, warning, timing), bodies (volume, area, centre of mass, bbox, face/edge/vertex counts, visible; removed_later: true while a sketch is being edited on a rolled-back timeline and a later feature deletes the body, which is then not shown), sketches (curves, points, constraints, profiles, DOF, solve status, visible), joints when the design has any (as joint.inspect gives them: each origin's occurrence, component and path, the snap as saved, the feature, face, edge or sketch curve (sketch: {id, name, curve}) its persistent names refer to, resolves and the error when it is lost; kind, offset, angle, flip, values, limits, expressions, motion links; execute joint.inspect {joint} gives one joint), selection, undo depth. Hide helper sketches, bodies or construction planes with execute browser.visibility ({items, visible} or {folder, visible}), and joint glyphs with {joints: [id or name], visible} or {folder: "joints", visible} (a hidden joint still constrains and drives; joint.list reports each joint's visible).
measure body? or items? Volume (mm³), area (mm²), centre of mass, bbox and topology counts per body, with totals. With items (one or two selections as execute takes them: face, edge, vertex, sketch_curve, profile), the picked geometry instead: see below.
body_topology body Edges (index, midpoint, length, ends) and faces (index, name, area, centroid, normal, point, and for a cylindrical face its cylinder: origin, axis, radius). An edge midpoint is the [x,y,z] that picks that edge for fillet and chamfer. A face's point lies on it and picks it (its centroid may not: that of a face with a hole is in the hole); a face's name picks it for the commands that take faces (face.delete, solid.shell, solid.draft, solid.offset_face, solid.split_face, solid.replace_face, and face of sheet.convert).
set_parameter name, value or expression (number or expression), unit?, comment? Create or change a user parameter and recompute. expression is another name for value (the one parameters.change uses): give one of them, or both only when they agree; neither, or two that differ, is an error. Reports recomputed (features rebuilt), restored (features whose earlier result for the same inputs was reused, e.g. a value set back) and features that now fail.
screenshot view? (iso, top, front, back, bottom, left, right, home, fit, or a named view saved with view.save), camera? ({target, yaw, pitch, distance, fov?}, as view.apply), path?, width?, height?, source? (window, model), grid?, axes? A PNG image content block. Headless: a render with the viewport's pipelines on a headless GPU device (see solidcraft-cli gpu; with no adapter, the CPU rasterizer, without analysis colours), fitted to the model, or from the named view or camera; the reply's renderer says which (gpu, gpu-software, cpu, with a warning for cpu). Bridged: the app window (default) or a model render with the app's camera; a named view or camera moves the app's viewport there first (execute view.apply does the same without a screenshot). grid: false and axes: false leave the ground grid (it lies on Z 0 and cuts through a part that straddles it) and the origin's axes out of a model render; axes hidden with browser.visibility stay out whatever axes says. Bridged, a screenshot that gives grid or axes is a model render, which follows the app's Grid and Origin switches for what it does not give; source: "window" with either is an error. A window that is minimised or on another macOS Space presents no frame: with no source the reply is a model render and says so (fallback); source: "window" is an error saying why, at once. Agents that don't need the app's chrome can ask for source: "model".
export path, format? (step, 3mf, 3mf-tagged, stl, stla, obj, urdf), bodies?, mesh_prefix?, effort?, velocity? STEP, 3MF or mesh export; URDF writes the whole assembly with its meshes next to it (docs/urdf.md).
undo, redo — Undo / redo one change.
pending — Bridged: the result of a request that was still running when its call returned (an error starting still running:), waiting up to 60 s for it; done: false while it runs (call again), pending: null when nothing runs. See Long-running requests.
crash_reports id?, limit? The crash reports on this computer, newest first; with id, the whole report (backtrace, command, recent commands, design_path/replay_path). See crash-reports.md.
skill_reference skill, file? (default reference.md) Read a bundled skill document. Skills: solidcraft-modeling, solidcraft-mechanisms, solidcraft-editing, solidcraft-checking.
running - What the app is running or has queued, for every client (control.running, see control-protocol.md): each request's connection and id, method, command id or a batch's step count, running_s and queued, never its parameters. Answers at once while the app is busy: ask it when a call timed out before sending a change again. Headless: always {"running": []}.
new_design name? Start an empty design.
open path Open a .solidcraft design, a STEP file (.step/.stp) as a new design holding its bodies (an Import base feature), or a 3MF/STL file as mesh bodies.
save path? Save the design (to its current file without path).

measure with items reports each item under items (length_mm, area_mm2 or position). A circle, arc, hole or other cylindrical face also gives diameter, radius, center and axis, with circumference when it goes all the way round, or arc_length and angle_deg for an arc; round says which (circle, arc, cylinder). A slot (a sketch profile of two half circles and two lines, or one half-cylinder end of a slot in a solid) gives slot: width, center_distance and overall_length. With one item these fields are repeated at the top level. With two items:

  • min (also distance_mm, with from, to and delta): the closest distance;
  • max (with max_from, max_to): the farthest distance;
  • center_distance (with center_from, center_to): when an item is round, from its centre or axis: axis to axis for parallel axes, the closest distance between the axes otherwise, and axis to plane, line or point;
  • measure: the distance the Measure panel shows first (center where there is one, else min);
  • angle_deg between two planar faces or straight edges.

Earlier fields are unchanged, so older scripts keep working.

To change a feature after it is made, read its definition with execute timeline.inspect ({feature}: id or name) and change fields with timeline.edit ({feature, set}; e.g. {"extent": {"distance": "30"}}). The fields per feature type are listed in the book's Solid features chapter.

Resources: solidcraft://design (inspect JSON with measurements) and solidcraft://commands (the catalog).

Errors: an unknown tool or arguments that don't match the tool's schema are JSON-RPC -32602 errors (an unknown argument's error lists the tool's arguments); an engine failure (bad command parameters, a feature that fails) is a tool result with isError: true and the engine's message, which the model can read and correct.

Long-running requests

The app answers a control request that runs longer than 60 seconds (a big fillet, say) with "still running" (control-protocol.md); it has not failed, and was not cancelled. The bridged server then keeps the connection, where the reply will come, and asks the app (control.pending) every half second for up to 240 seconds more, saying on stderr how long it waits; when the request finishes, the tool call returns its real result. If it still runs after that, the tool call fails with an error starting still running: that names the request and says what it still does: it will still apply when it finishes, or will still write the file (an export or a save: the design does not change), or only reads (nothing changes). Don't send it again: call the pending tool, which waits up to 60 seconds for its result each time. Any other tool call first waits up to 60 seconds for it too, and is not sent while it still runs (so nothing is queued to apply unseen behind it); a request that finishes in that wait keeps its result for pending. If the connection to the app is lost meanwhile, the result is lost with it, and pending says so: inspect the design before sending a change again.

So a single tool call may block for up to 300 seconds (the app's 60 s wait plus 240 s of polling) before it returns still running:, and a call made while an earlier request still runs for up to 60 seconds more. An MCP client needs a per-call timeout above that (say 330 s) to see that error; a client that gives up sooner only sees its own timeout, though the request goes on running in the app: then call pending (or inspect_design) before sending anything again.

A batch step that is still running stops the batch like a failure. Its rollback first waits for the step (as above), so a restored checkpoint takes the step back too; if it still runs, the reply says rolled_back: false, marks the step still_running and lists it in applied, as it will apply when it finishes.

batch rollback goes back to an engine checkpoint (edit.checkpoint) taken before the first step: the design, the active sketch and the undo history are restored exactly as they were, however long the history is. Batches that call file.new or doc.open get the design back, not its file path. (An older app without checkpoints falls back to undoing as many steps as the batch added to the undo history.)

Camera and close-up recipe

The camera of screenshot (the object view.apply, view.save and solidcraft-cli snapshot --camera take too) is an orbit camera, {target, yaw, pitch, distance, fov?}, in millimetres and radians. Unlike the angle of solid.move, a bare number is not degrees here. The eye is at target + distance * (-sin(yaw) * cos(pitch), -cos(yaw) * cos(pitch), sin(pitch)), looking at target with Z up on the screen. So yaw: 0, pitch: 0 is the front view: the eye at -Y, looking along +Y, with +X to the right. A yaw of π/2 looks from -X, π from +Y and -π/2 from +X. The pitch is the elevation angle up from the XY plane, and π/2 looks straight down with +X to the right and +Y up. fov is the vertical field of view in radians. 0, or no fov, is an orthographic view whose visible half-height is 0.4 × distance; with a perspective fov the half-height at the target is distance * tan(fov / 2). inspect.section removes what lies on the side its normal points to: it keeps the points p with n · p < d, where d = n · origin (flip: true keeps the other side). To look into the cut, put the eye on the removed side.

To look closely at one part of a large model, such as a STEP assembly, work in a headless session. These steps change only the open design in memory:

  1. Open the model with solidcraft-cli mcp --in assembly.step (or the open tool, which also takes up and names, see step 4 and STEP import). inspect_design lists the bodies with their names and bounding boxes.
  2. Remove what is in the way with solid.remove, e.g. {"bodies": ["Cover"]}.
  3. Cut the region out of the rest with solid.box, e.g. {"corner": [40, 0, -45], "length": 20, "width": 50, "height": 45, "operation": "intersect", "targets": ["Housing", "Plate"]}. Each listed body is trimmed to the box, and one that lies wholly outside the box is removed. The box itself is not kept.
  4. If the model lies on its back (many STEP files have +Y up), open it with up: "y" (the open tool, doc.open, file.insert_step), which turns the file's bodies a quarter turn about X (Y becomes Z). One already open is stood Z up with solid.move, listing every body you keep, e.g. {"bodies": ["Housing", "Plate"], "axis": [1, 0, 0], "angle": 90}. It turns the bodies about the world origin, and a bare angle is in degrees. A turn of +90° about X takes +Y to +Z.
  5. Colour the parts that matter with appearance.assign, e.g. {"bodies": ["Housing"], "color": "#3a7bd5"}, and hide helper bodies or sketches with browser.visibility, e.g. {"items": ["Plate"], "visible": false}.
  6. Cut a section with inspect.section, e.g. {"plane": {"origin": [50, 30, 0], "normal": [0, 1, 0]}} or {"plane": "XY", "at": 5}.
  7. Take the picture with screenshot: a camera whose target is the centre of the region and whose eye is on the cut side, source: "model", grid: false if the region straddles Z 0 (the ground grid would cut through it), and path, width and height as needed. An orthographic picture is 0.8 × distance high, so a distance of about 1.5 times the region's size frames it. Look at the image, adjust yaw, pitch and distance, and shoot again. Headless, the picture is always a render of the model; bridged to the app, source: "model" asks for that render instead of the app's window.

Compact replies

execute and batch default to verbose: true, preserving the full command results. For long modelling sessions, set verbose: false at the tool-argument level:

{"name": "execute", "arguments": {"command": "solid.box", "params": {"length": 10, "width": 10, "height": 10}, "verbose": false}}

Feature replies omit the full bodies list while keeping feature, name, created_bodies, changed_bodies and removed_bodies. A cut can therefore report a changed body even when no body name was added or removed. Sketch replies keep new IDs under created and omit only identical top-level points, curves, constraints and dimensions lists. Other fields and errors are unchanged; inspection commands still return their complete results. Compaction uses the engine's reply directly, without an additional design-inspection request.

Demo: an enclosure built over MCP

examples/demo/enclosure_mcp.py is a small MCP client. It drives a running app (solidcraft --control PORT) through solidcraft-cli mcp --connect and builds an electronics enclosure one execute call at a time (enclosure_steps.json): a rounded box, shelled open at the top, a USB port cut through a side wall, four screw bosses, and then a change of the width parameter that the timeline rebuilds. examples/demo/record.sh records the run with ffmpeg, with each step's caption burnt in.

examples/demo/enclosure_ui_mcp.py (v2) builds the same part the way a person would: the modelling goes through the app's UI over its control channel. The pointer glides to toolbar buttons and faces and clicks them, the real dialogs open, the sketches go on the model's faces, and the camera follows. The parameters, the sketch dimensions bound to them and two extrudes go over MCP. With --story v3 (the default) one standoff is patterned 2 × 2 in the Rectangular Pattern dialog with spacings from the parameters, and width and wall are then changed in the Parameters dialog. record_ui.sh records it and draws the pointer and the captions.

A typical session

→ {"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"me","version":"1"}}}
→ {"jsonrpc":"2.0","method":"notifications/initialized"}
→ {"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"batch","arguments":{"commands":[
     {"command":"sketch.create","params":{"plane":"XY"}},
     {"command":"sketch.rectangle.two_point","params":{"p0":[0,0],"p1":[40,30]}},
     {"command":"sketch.finish"},
     {"command":"solid.extrude","params":{"distance":20}},
     {"command":"solid.fillet","params":{"edges":[[0,0,10]],"radius":3}},
     {"command":"sketch.create","params":{"plane":"XY"}},
     {"command":"sketch.circle.center","params":{"center":[20,15],"radius":5}},
     {"command":"solid.extrude","params":{"distance":20,"operation":"cut"}}]}}}
← {"jsonrpc":"2.0","id":2,"result":{"content":[{"type":"text","text":"{\"ok\": true, \"completed\": 8, \"results\": [...]}"}]}}
→ {"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"measure","arguments":{}}}
← … "total": {"volume_mm3": 22390.6, …}
→ {"jsonrpc":"2.0","id":4,"method":"tools/call","params":{"name":"screenshot","arguments":{"view":"iso"}}}
← {"jsonrpc":"2.0","id":4,"result":{"content":[{"type":"image","mimeType":"image/png","data":"iVBORw0…"},{"type":"text","text":"{\"source\":\"model\",…}"}]}}
→ {"jsonrpc":"2.0","id":5,"method":"tools/call","params":{"name":"export","arguments":{"path":"/tmp/part.step"}}}

The tests in crates/mcp/src/tests.rs run sessions like this (box → fillet → hole, parameters, batch rollback, hostile arguments) and check the results only through MCP.