Two review findings, both root causes.
1. `_sync_header_to_dimstyle` copied the DIMSTYLE entry's ezdxf-internal
arrowhead spelling (OBLIQUE) into the DXF-facing $DIMBLK/$DIMBLK1/$DIMBLK2
header, so the saved file named a block it did not hold (the blocks are
_OBLIQUE / _DOT) and `system_get_variable("DIMBLK")` answered "OBLIQUE"
headlessly against "_OBLIQUE" on COM and in the same engine's own
`dimstyle_list` row. The header write now goes through
`reported_arrowhead` — the same read-side rule `_dimstyle_value` applies.
Pinned by header, `system_get_variable`, save/reopen (`ezdxf.readfile`
header equals the block the file holds) and the Restore-to-Standard
empty-string case in `test_builtin_arrowheads_render_save_and_report_one_spelling`.
2. The `mleaderstyle` capability refusal rested on a false engine boundary.
acax25enu.tlb (AutoCAD 2026) exposes IAcadMLeaderStyle with Name,
ArrowSize, LandingGap, TextHeight, TextStyle, ... — exactly what
`Dictionaries.Item("ACAD_MLEADERSTYLE").Item(i)` returns — and
IAcadDictionary.AddObject(keyword, "AcDbMLeaderStyle") creates one.
COM `mleaderstyle_list` now reads the four values off those objects
(`values_available: True`); COM `mleaderstyle_create` validates every value
(`validate_mleaderstyle`, shared with ezdxf), checks the clash and the text
style before AddObject, then writes the four properties. The contract
method is `@abstractmethod` again, the key is gone from both capability
maps, the fake ActiveX surface models the real call shape, and the spec
(§4.2, §8.1, §9) states the ActiveX route.
Live AutoCAD was not running on this machine (GetActiveObject: operation
unavailable; no acad.exe), so the COM route is type-library-verified and
fake-tested, not executed live — scripts/smoke_settings_com.py (Task 24)
owns that.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Two blocking review findings on the standards data layer:
* DIMTP and DIMTM were pinned to ("float", 0.0, 1e6), refusing every signed
deviation with a false "must be at least 0". AutoCAD documents both as
signed and the repo already relies on it: build_dim_override writes
dimtm=-0.022 for ISO 286 p6 and dimtp=-0.020 for f7, and
dimension_linear(fit=...) emits exactly those. A dimension style could
therefore never hold a double-positive or double-negative fit that the same
server draws correctly per dimension, and an agent obeying the message
would ship a sign-flipped tolerance. Both are now ("float", -1e6, 1e6); a
provenance test resolves the p6/f7/H7/h6 overrides build_dim_override
emits through resolve_dimstyle and pins the signs.
* 827719f made check_dim_value return DIMBLK/DIMBLK1/DIMBLK2 in the
canonical _OBLIQUE spelling and added ezdxf_arrowhead without telling
Task 10, whose plan helpers store the value verbatim and read the raw
string back - executed as written the headless drawing raised at render
and at doc.write, a user block passed the data layer and failed at save,
and ezdxf's OBLIQUE-in-session / _OBLIQUE-after-reload split made an
idempotent re-set report a change (spec section 8.3). This commit adds the
read-side twin reported_arrowhead (never raises; a foreign name passes
through), pins it with an in-session-equals-reloaded round trip, and edits
the plan: Task 9 Produces and Task 10 Consumes now name ARROWHEAD_BLOCKS,
canonical_arrowhead, ezdxf_arrowhead and reported_arrowhead; Task 10's
_write_dimstyle_values maps through ezdxf_arrowhead, _dimstyle_value and
_com_dimvar_for_report map through reported_arrowhead, a new
_require_arrowhead_blocks refuses a user block absent from doc.blocks
before dimstyle_create/modify write anything (and rewrites the name to the
block table's own spelling, measured to be what a reload reports), and two
headless tests witness the rendered INSERTs, the save/reopen spelling and
the refusal. The plan's own copy of the range table is corrected to match
the module (arrow kind, signed tolerances). The edited helper block was
executed against ezdxf 1.4.4 before being written into the plan.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Skeleton (contracts, interfaces, file ownership, merge notes) plus the
drafted task steps for Track A hardening (H), styles (S), settings (C),
page setup and templates (P), environment (V) and the final tasks (F);
25 cross-group consistency findings reconciled before commit.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
38 tools in three sections: ISO-25/ANSI dimension, text and leader styles
with drawing_apply_standard; drawing_settings growth, a system-variable
catalogue and document properties; page setup with PDF mediabox read-back,
batch plot and five bundled templates built by the server's own tools;
multi-document on both engines, portable layer states, named views, UCS,
launch, preferences and operator prompts. Track A hardening is wave 0.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
scripts/smoke_pid_com.py runs the whole track against the live AutoCAD in
a new document. ComBackend.connect() is lazy by design, so the script reads
reachability from system_status()["connected"] rather than from connect()
as the plan sketched, and closes its document without saving unless --keep
is passed. On this machine (AutoCAD 2026, "25.1s (LMS Tech)") it exits 0:
5 nodes, 4 edges, 0 dangling, confidence_min 1.0, 0 critique issues,
2 XDATA chunks — identical to the headless engine on the same spec. The
first attempt exited 2 on a stale win32com gen_py cache (a package folder
holding only __pycache__); an environment fault, no backend change needed.
A/B against v1.5.1 published as ab-v1.5.1-vs-v1.6.0-dev.json: 26/29 ->
29/29, three miss -> pass, zero regressed. check_doc_numbers.py now names
the report and the two row labels in one place (AB_REPORT, AB_*_LABEL),
and README.md / benchmarks/README.md carry the new rows; the v1.5.0 and
v1.4.0 tables stay as history.
README: P&ID row, the showcase and catalogue images, the token table
re-measured (161 / 44,930 by default, 152 / 41,420 with TOOL_PACKS=core,
lean 50 / 14,007), every stale count reconciled (154 -> 166 registered,
149 -> 161 advertised, corpus 158/672 -> 162/731 by the same count, 19 ->
20 groups, 6 -> 8 resources, six-task matrix with pid_roundtrip), a
"Known limitations of the P&ID track" subsection, and the snapshot line at
this machine's 2423 collected tests (CI has not run on this tree; noted in
the changelog with the pre-existing 1369 vs 1363 drift).
CLAUDE.md: section 17, sections 5 and 7 counts, TOOL_PACKS, the two 1.6
contracts and the reader rules. Spec status -> implemented. The plan and
spec markdown were reformatted by ruff (fenced Python only) because
`ruff format --check .` was already red on them at HEAD and the gate
requires the whole tree clean.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Two authoring slips in the valve x actuator composition:
- The diaphragm actuator's stem stopped at y=4 while the dome's flat base
sat at y=6, so the canonical control-valve symbol (PID_VALVE_*_DIAPHRAGM,
9 catalogue entries) shipped with its dome floating 2 mm above the stem.
- Actuators were placed at a fixed y=4 regardless of the body, so on the
needle body (top 5.6) the piston/motor/solenoid glyphs were drawn over
the knob and the hand bar across the needle's stem; the stem line(0,0,0,4)
also duplicated the needle's own axial line (and the butterfly disc).
Placement is now body-relative: `_body` reports `axis_top` (where its own
geometry on x=0 ends) and the stem runs from there to the glyph base at
`body_top + STEM_CLEARANCE`; the diaphragm dome's base is the stem top
(base + DIAPHRAGM_YOKE), and the FAIL attribute follows the glyph base.
The reference 8x4 bodies keep their published numbers (globe+diaphragm
signal at y=9.0, bbox top 9.0).
Tests: per-combination check that the stem starts at the body's derived
axis top, the glyph's lowest edge equals the stem top, the glyph clears
the body, and the signal port is the topmost edge; plus a catalogue-wide
no-duplicate-primitives gate. 33 combinations fail on the old geometry.
The plan's Step 3 snippet is synced with the fixed module.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Plan covers the prerequisite block_insert attribute fix (ezdxf wrapped the
INSERT in an anonymous *U block and lost its ATTRIBs), the two generic
contracts, the 43-symbol catalogue, routing, the ISA-5.1 grammar, the
graph reader, deliverables, six critique focuses, pid_from_spec, TOOL_PACKS,
resources, benchmark v4, and the live COM smoke.
Spec reconciled with the plan: 43 base symbols, FUNC/LOOP at +-2.0,
solid = DXF SOLID entity, shape/direction options, example spec uses an
actuated FCV-101 (a hand valve has no signal port), refine hints in
detail["hint"], preflight LAYER_SET_INTENT_MISMATCH, radial port matching
on the circle.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
Research snapshot of the AutoCAD MCP field (20 READMEs, 2026-09-14), the
seven 1.6 tracks with measurable acceptance criteria, their order
(A -> E -> B -> G -> F -> H -> D), and the tool-pack decision.
Track A: symbol catalogue authored in code (ISO 10628-2 + ISA-5.1), two
generic contracts (block_define, entity_get/set_xdata), nine pid_* tools,
routing, the ISA-5.1 tag grammar, the pid_graph reader, six critique
focuses, TOOL_PACKS, and the test/evidence plan.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd
The triage over the 43 unaddressed verdicts returned 4 code and 35 doc items.
The code ones shipped in e24b46d; these are the documents.
Twenty-one applied mechanically after a validation pass confirmed each locator
still matched exactly once and each replacement was supported by the tree. Ten
were held back because a re-verify pass found them mutually exclusive -- three
patches rewriting the same benchmarks/README block, three rewriting the same
OCS paragraph -- and applying them in sequence would have corrupted the files.
Those were resolved by hand into one coherent edit each. One was rejected.
The corrections worth naming:
- **The A/B table contradicted itself in three files.** benchmarks/README.md
and CLAUDE.md said 24 checks / 21-24 / 87.5% / "fixed 3" while the suite has
26 and the published JSON says 21/26, 80.8%, five fixed. All three now agree,
and the five are split into the three new-capability `miss -> pass` and the
two repaired-defect `fail -> pass`, which is the distinction the lane exists
to make.
- **"Every number below is produced by a script"** was false twice over:
section 5's nine scores are literals a human assigned, and section 4 now
quotes hand-collected timings. The claim is narrowed to the charts, which
really are all rendered from committed data, and both exceptions are named.
- **The OCS paragraph over-quantified.** TEXT rotation was given as the single
exception; z components and HATCH boundaries are also entity-frame, and the
live backend records `2d_polyline_coordinates_not_verified_against_live_autocad`
in the same reason string. It now points at `ocs_normalized` as the live list.
- **Transactions were shown as bare parity.** The server itself grades them
differently -- `FeatureCapability(True, "undo_mark")` on COM against
`(True, "snapshot")` headless -- and COM rollback sends `_UNDO B` without
confirming it landed. The row says so.
- **"Raw commands: Opt-in"** named a switch that does not exist. Both tools are
registered unconditionally; `DANGEROUS_COMMANDS_ENABLED` only disables the
sanitizer.
- **"Wave A shipped 13 of a planned 39"** hardened an approximation: the only
source writes `~39`.
- **CAD_PROGID** said nothing beyond the connection was tested against
BricsCAD/ZWCAD/GstarCAD, implying the connection was. It was not.
Perf lane reworked against a fresh paired measurement. The old text claimed
"28% slower than v1.4.0, median of three runs" and that
`EZDXF_CALL_TIMEOUT=0` restored v1.4.0's numbers "exactly". Three problems:
`perf_suite` runs each workload once per invocation and cannot produce a
median, "exactly" was refuted by the figures in its own sentence, and the
number moved when the timeout wrapper was rebuilt. It now reports the guard's
cost -- +20.8% and +19.9%, three paired runs, median taken by hand -- says the
figures have no published artifact, and says the wrapper rework recovered about
a third of the v1.5.0 overhead rather than closing it.
perf-ezdxf.json and the perf chart regenerated against the shipped code, and
the README table rebuilt from that artifact so the two cannot drift.
1267 passed, 1 xfailed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Three defects found after 1.5.0 reached PyPI, two of them by the adversarial
pass over the rewritten README.
The HTTP bind guard failed open on the launch path it was written for.
`_guarded_run_async` read the bind address from the `host` kwarg and treated an
absent one as loopback. It is not: `fastmcp run` forwards that kwarg only when
`--host` was passed, and fastmcp otherwise resolves the host from
`fastmcp.settings.host`, which is environment-backed. So
FASTMCP_HOST=0.0.0.0 fastmcp run server.py:mcp --transport http
bound every interface -- 149 tools including file open/save and AutoCAD command
execution, no token -- while the guard inspected the literal "127.0.0.1" and
waved it through. v1.4.0's changelog claimed this guard "fires on every launch
path"; it fired, and looked at the wrong host. It now resolves the effective
host, and warns rather than staying silent if it cannot.
`[com]` and `[full]` could not be installed off Windows. Both listed
`pywin32>=306` with no environment marker, and pywin32 publishes Windows wheels
and no sdist, so `pip install autocad-mcp-pro[full]` died at resolution on
Linux and macOS -- and `[full]` is the extra the README recommended to
contributors. CI never caught it because every Linux lane installs `[pdf]` or
syncs from uv.lock, whose resolution markers hide the gap. Both entries carry
`; sys_platform == "win32"` now, with a release-consistency test to keep them.
`drawing_settings({"dimscale": ...})` reported success and changed nothing
headlessly. It writes `$DIMSCALE`, which is what AutoCAD reads when it draws a
dimension, so the call worked on COM; ezdxf renders from the dimstyle table
entry and never consults the header. One call, one reported success, two
different drawings. The header value now folds into the per-dimension override
on all five dimension tools -- including `dimension_aligned` and
`dimension_angular`, which do not share the tolerance-override path and the
obvious fix would have missed. DIMSCALE 1.0 writes nothing.
README: every engine-conditional family in the feature table now says which
engine (boundary tracing, WIPEOUT/REVCLOUD/MTEXT masks, CHSPACE, typed hatch
edge paths, headless undo), and the checkout install line stopped recommending
an extra that could not resolve on the reader's platform.
1257 passed, 1 xfailed. 1.5.0 is to be yanked: its diameter and radius callouts
put wrong numbers on drawings at default settings.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
`add_diameter_dim` / `add_radius_dim` measure to the point they are handed;
`leader_length` is a text placement. The ezdxf backend passed
`centre + (radius + leader)` as that point, so the leader became part of the
measurement: a true 40 mm bore was dimensioned 60 at the default
`leader_length=10`, and 108 at 34. Radius callouts were `leader_length` too
large the same way.
This is the worst class of defect for this project -- a wrong number, at
default settings, on the one part of a drawing that exists to carry numbers --
and it shipped in v1.4.0 and earlier. The size now comes from `radius` and
`location` only places the text.
The live COM backend was never affected: ActiveX's AddDimDiametric takes the
two chord points and the leader length as separate arguments. The two engines
had been disagreeing by 2 x leader_length on the same call.
Found while building the drawing for the new README, whose own bore callout was
wrong. Two checks added to correctness_suite.py; both report fail -> pass
against v1.4.0, not miss -> pass, which is what separates a repaired defect
from new capability in that lane. A/B is now 21/26 -> 26/26, five fixed, zero
regressed.
README rewritten from scratch
- Hero is a real ISO A3 sheet the server drew: m6 z24 involute spur gear, DIN
6885 keyed bore, section A-A, ISO 129 dimensions, ISO 286 H7 fit, ISO 7200
title block, 56 entities, drawing_critique clean. New
scripts/render_readme_showcase.py rebuilds it, and the tool-call sequence
that produces it is printed in the page.
- Star history chart, badge row, capability matrix, five evidence lanes, and a
"what this release is bad at" section carrying the 28% throughput regression.
- Every image and repo link is now an absolute URL: README.md is the PyPI long
description, where relative paths 404.
Corrections this work forced, all measured
- The df=0 example was wrong. FILLET is already in entity_fillet's own text;
the commands that genuinely scored zero are WBLOCK, PEDIT, OVERKILL,
LAYTRANS and MATCHPROP.
- 40,305 tokens is the tools/list wire payload; ~25,143 is the part forwarded
to the model. The old wording said "of schema", which is neither.
- `--tokenizer anthropic` needs a package and credentials this repo does not
lock, and refuses rather than falling back. Said so.
- Discovery ranks over the 149-tool advertised catalog, not the 154 registered.
- Five middlewares, not four; the fifth is what makes refusals machine-readable.
- Five of the 27 capability keys are environment-dependent, so the matrix
carries footnotes instead of bare ticks.
- "Every coordinate is WCS" has one deliberate exception: TEXT rotation.
- render_live_chart's axis still said "fixed 10-task matrix" while our bar came
from v3's fifteen. Each bar now carries its own matrix.
- .env.example said MAX_UNDO_STACK=50 (code: 5) and omitted four real
variables, two of them the remote-HTTP controls.
- server.json described an empty ALLOWED_PATHS as blocking access. It disables
the allowlist.
- server.py called the lean profile ~46 tools; it is 47.
- backends/capability.py cited a test file that does not exist.
1249 passed, 1 xfailed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Tool count 131 -> 154; 474 -> 1242 collected tests.
The release exists for two measured reasons. A 149-tool catalog costs a client
~40,300 tokens before it has asked for anything, and a drafter searching the way
drafters think -- FILLET, BPOLY, QSELECT, WBLOCK -- scored df=0 against the stock
index, because those words were simply not in the data.
Discovery and batching
- DISCOVERY_MODE=search replaces the catalog with search_tools + call_tool,
ranked over an AutoCAD command and synonym corpus (discovery/aliases.py) on
top of fastmcp's BM25 index. Measured: 40,305 -> 356 tokens advertised.
- cad_batch runs a step list in one round trip; fields= projects result-heavy
responses. The `core` profile is gone -- discovery replaced it.
Honesty debt (Tier 0)
- OCS/ECS: `extrusion` appeared nowhere in this codebase, so mirrored and
rotated geometry reported coordinates that were quietly wrong. backends/ocs.py
translates at the backend boundary; every tool coordinate is WCS on both
engines, and a tilted plane refuses rather than answering with a number that
looks fine.
- drawing_save_as(fmt="dwg") wrote DXF bytes to a .dwg path. drawing_audit never
repaired. analysis_measure_area shoelaced points the caller typed in.
- ezdxf calls gained a per-call timeout, so one hung call can no longer wedge a
server whose document lock is a single asyncio.Lock.
Architecture
- The 1854-line backend ABC is 19 contract modules; @capability(key, reason=)
makes optional methods declarative and a gate fails on undeclared holes.
- 8 layout/viewport tools, CHSPACE, CAD_PROGID escape hatch.
- Wave A: selection filters, boundary tracing, hatch depth, annotation objects,
measurement -- 13 tools of a planned 39. The REGION and 2D-boolean family was
cut on evidence: add_region() yields zero ACIS bytes and the greiner_hormann
substitute loses 28.2% of the area on a square with one semicircular edge.
Verified against a live AutoCAD 2026, which found four COM defects no amount of
reading would have caught -- AcadPViewport has no ViewCenter, and the line that
set it had been swallowed by a bare except since v1.4.
Benchmark v3 (15 tasks) and what it caught
- v2 scored this server 10/10, which carried no information: every task in it
exercised something the server was built around. Five tasks were added because
they can fail, and three did while being written -- a hatch could not be
measured though the refusal listed HATCH as measurable, the boundary tools and
the measure tool disagreed about the same shape (139.2177 vs 139.2699), and a
circle stored as two bulged vertices measured 0.0.
- Competitor reports are NOT re-scored. They are pinned v1.4 runs against v2 and
were never asked these five questions; the chart shows those rows `not run`.
- Correctness A/B: 21/24 -> 24/24, zero regressions.
Performance, stated plainly
The published v1.4 perf report was recorded on CPython 3.14 and the new one on
3.11.15, so the apparent 2-6x speedup is the interpreter, not this release.
Re-measured on one interpreter, v1.5.0 is 28% slower creating entities and 22%
slower on the 10k roundtrip. Fully attributed: EZDXF_CALL_TIMEOUT=0 restores
v1.4.0's exact numbers. That is the deadlock fix being paid for, the knob is
documented, and winning it back is on the 1.6 roadmap.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Leaderboard restyled to the LLM-benchmark look: dark background, rank-numbered labels, per-rank color palette, 'NN / 100 - grade' annotations. The live-run, task-matrix and performance charts share the same dark theme for a consistent README identity. Charts regenerated deterministically from their data files.
Co-authored-by: Cursor <cursoragent@cursor.com>
source_review.json extended with three projects reviewed 2026-07-23 (LokmenoWer/best-cad-mcp B/74, daobataotie/CAD-MCP B/35, zh19980811/Easy-MCP-AutoCad C/30) and our entry updated to the released v1.4.0 (95: CI matrix, PyPI packaging, A/B gate, paper space, ISO 286, opt-in solids). Chart regenerated via render_chart; README benchmark section now opens with the leaderboard + 9-row table, source-review lane merged into it.
Co-authored-by: Cursor <cursoragent@cursor.com>
