Skip to content
Back to search

agentcad

CAD CLI and MCP server for AI agents

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today

Check it deeper on your machine

It has changed since its scan on 11 Nov 2026.

As a Claude Code plugin, install it and then ask Claude Code:

/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/jdilla1277/agentcad with scanskillsafety

Or from a terminal, clone it and run it:

git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/jdilla1277/agentcad'

It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.

Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/jdilla1277--agentcad/badge.svg)](/listings/jdilla1277--agentcad)

HTML

<a href="/listings/jdilla1277--agentcad"><img src="/api/listings/jdilla1277--agentcad/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

Install agentcad, then paste this into Claude Code, Cursor, or any coding agent:

Create a Python 3.12 virtual environment, then:

pip install agentcad
agentcad init --name phone-stand

Design me a phone stand: a simple angled cradle that holds a phone
at 60 degrees. About 80mm wide, 50mm deep, with a 5mm lip at the bottom
to keep the phone from sliding. Show me a preview when you're done.

Keep generated files in a build directory

Create `agentcad.toml` in your source project before initialization:

build_dir = "./build"

Then run `agentcad init`. Version directories, the generated manifest, viewers, exports, and logs go under `build/`; authored scripts and installed guidance stay in the source project. Add the build directory to your own `.gitignore`.

For CI, initialize and select an independent root with `agentcad init --build-dir /tmp/cad-build` and `agentcad run model.py --label first --build-dir /tmp/cad-build`. Overrides do not edit project configuration. Relative build paths resolve from the project root, not the caller's working directory. Existing projects keep their current layout unless configured. `--output` remains a deprecated version label, never a destination. See `agentcad docs artifacts` for the full contract.

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

README.md · README · 451 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

This file has changed since it was scanned, so this is not exactly the text that was scanned, and no line is marked. The findings are listed with the safety scan above.

  1. # agentcad
  2. **CAD tool for AI agents.** Give your coding agent the ability to design 3D models.
  3. Your agent writes build123d Python scripts by default. agentcad handles execution, STEP export, PNG rendering, mesh export (STL/GLB/OBJ), geometric metrics, validation, diffing, and browser preview. CadQuery remains available as an explicit compatibility mode. Each command's final response is structured JSON on stdout.
  4. > **Reading the output:** the JSON response is written to **stdout**; human-readable progress and diagnostics go to **stderr**. Parse stdout as JSON and treat stderr as plain text — don't merge the streams with `2>&1` before a JSON parser, or the progress lines will break parsing. If you need both, capture them separately.
  5. agentcad is open source under the Apache License 2.0. It runs locally and requires no signup.
  6. [![Featured on Product Hunt](https://api.producthunt.com/widgets/embed-image/v1/featured.svg?post_id=1165633&theme=light)](https://www.producthunt.com/products/agentcad?utm_source=badge-featured&utm_medium=badge&utm_campaign=badge-agentcad)
  7. ## Demo
  8. [![Watch a coding agent design in agentcad](https://img.youtube.com/vi/Zsn31-IilWM/maxresdefault.jpg)](https://www.youtube.com/watch?v=Zsn31-IilWM)
  9. A coding agent designing in agentcad, live. See more at [agentcad.dev](https://agentcad.dev).
  10. ## Recently from agentcad
  11. - **November 11, 2026 — [From PDFs, images, and drawings to Revit](https://agentcad.dev/revit):** Model from drawings, photos, and measurements, review the geometry, and bring checked STEP files into Revit.
  12. - **November 6, 2026 — [v0.7.2: Retry builds without creating duplicate versions](https://agentcad.dev/changelog/v0.7.2):** Recover saved results after a lost response, keep projects separate, and preserve non-English text on Windows.
  13. - **October 28, 2026 — [Measuring agentcad's effectiveness, again](https://agentcad.dev/measuring-agentcad-effectiveness-again):** The workflow candidate behind v0.7.0 scored 33% higher overall than the previous agentcad result, with editing up 42% and 80 of 81 valid outputs.
  14. - **September 29, 2026 — [Using Meta's Muse agent for 3D printing and CAD](https://agentcad.dev/muse-3d-printing):** A CAD workflow from a phone with Meta's Muse agent, including a video walkthrough.
  15. - **September 3, 2026 — [The original effectiveness benchmark](https://agentcad.dev/measuring-agentcad-effectiveness):** The earlier comparison against direct build123d, with its [video walkthrough](https://www.youtube.com/watch?v=IacqGZH8Gu4).
  16. Browse the [change log](https://agentcad.dev/changelog) for more updates.
  17. The five-stage workflow below was evaluated in the v0.7.0
  18. [Candidate L CADGenBench run](https://huggingai4engineering-cadgenbench.hf.space/reports/james-dillard_gpt-5-nano-candidate-l-pi-harness-five-s_20261001-001100.html).
  19. GPT-5 Nano scored .245 overall (.175 generation, .353 editing) with 80 of 81
  20. valid outputs. This was a single release-candidate run, not an evaluation of
  21. the current release or a fresh matched direct-build123d comparison; see the
  22. [experiment closeout](docs/product-plans/agentcad-performance-next-steps.md)
  23. for the baseline, caveats, and next evaluations.
  24. ## Quick start
  25. Install agentcad, then paste this into Claude Code, Cursor, or any coding agent:
  26. ```
  27. Create a Python 3.12 virtual environment, then:
  28. pip install agentcad
  29. agentcad init --name phone-stand
  30. Design me a phone stand: a simple angled cradle that holds a phone
  31. at 60 degrees. About 80mm wide, 50mm deep, with a 5mm lip at the bottom
  32. to keep the phone from sliding. Show me a preview when you're done.
  33. ```
  34. To install the reusable agent skill separately, use
  35. [`npx skills add jdilla1277/agentcad-skill`](https://skills.sh/jdilla1277/agentcad-skill/agentcad)
  36. or [`clawhub install agentcad`](https://clawhub.ai/jdilla1277/skills/agentcad).
  37. Both install guidance, not the CAD executable; install `agentcad` with pip as
  38. shown above. For a project-local guide, run `agentcad skill install` instead.
  39. ### Keep generated files in a build directory
  40. Create `agentcad.toml` in your source project before initialization:
  41. ```toml
  42. build_dir = "./build"
  43. ```
  44. Then run `agentcad init`. Version directories, the generated manifest, viewers,
  45. exports, and logs go under `build/`; authored scripts and installed guidance
  46. stay in the source project. Add the build directory to your own `.gitignore`.
  47. For CI, initialize and select an independent root with
  48. `agentcad init --build-dir /tmp/cad-build` and
  49. `agentcad run model.py --label first --build-dir /tmp/cad-build`.
  50. Overrides do not edit project configuration. Relative build paths resolve from
  51. the project root, not the caller's working directory. Existing projects keep
  52. their current layout unless configured. `--output` remains a deprecated version
  53. label, never a destination. See `agentcad docs artifacts` for the full contract.
  54. ## Required agent workflow
  55. `agentcad init` installs a guide that requires the same workflow for new models
  56. and edits:
  57. > **Inspect → Specify → Build → Validate → Deliver**
  58. - **Inspect:** understand drawings, requirements, orientation, dimensions, and
  59. any source geometry before writing a model.
  60. - **Specify:** record measurable requirements and assumptions. For edits, also
  61. identify the target, expected change, and geometry that must remain stable.
  62. - **Build:** create the simplest geometry that can satisfy the specification;
  63. use the smallest bounded operation for an edit.
  64. - **Validate:** inspect visual, dimensional, difference, and structural evidence.
  65. If a check fails, return to the earliest incorrect stage and iterate.
  66. - **Deliver:** hand off only the exact STEP and requested exports that passed.
  67. ## Capabilities by workflow stage
  68. ### Inspect
  69. - **`agentcad import supplier.step --init`** — import a versioned baseline with
  70. a bounded inspection report, geometry IDs, and sampled four-view images.
  71. Retrieve the complete inventory or saved artifacts through `inspect`; no
  72. experimental flag is required. See the [artifact retrieval guide](docs/inspection-artifacts.md).
  73. - **`agentcad inspect output.step`** — bounded topology, orientation, addressable
  74. IDs, edit risk, and structural evidence
  75. - **`agentcad measure output.step`** — dimensions, edge lengths, face areas, and
  76. circular/cylindrical feature measurements
  77. - **`agentcad render output.step --view top,front,right,iso`** — visual evidence
  78. for orientation and feature location
  79. - **`agentcad parts list REF` / `parts show REF ID`** — inspect named parts and
  80. isolate one versioned part
  81. ### Specify
  82. - **`spec.json`** — records supported cylindrical-feature and structure
  83. requirements before modeling
  84. - **`agentcad check-spec output.step spec.json`** — evaluates that contract
  85. during validation
  86. ### Build
  87. - **`agentcad run script.py --label label`** — execute a build123d script,
  88. producing a versioned STEP file and geometric metrics
  89. - **Pre-injected helpers** — build primitives, import existing STEP geometry,
  90. select features, and perform bounded or validated editing operations
  91. - **[CAD recipes](recipes/)** — reusable features and models to copy and adapt,
  92. starting with a PCB enclosure with board supports and connector openings
  93. ### Validate
  94. - **`agentcad run ... --preview`** — four-view PNG for visual verification; the
  95. browser viewer can export an on-demand turntable GIF
  96. - **`agentcad diff 1 2`** — compare versions, including actual
  97. shared/reference-only/candidate-only source-frame volume for valid closed solids
  98. - **`agentcad view old.step new.step`** — open a synchronized A/B comparison
  99. with separate centered projection and source-frame 3D volume artifacts
  100. - **Live project viewer** — successful runs open one stable local project URL
  101. and refresh it after subsequent completed builds. Leave the tab open while
  102. iterating; camera and compatible review settings survive updates. Failed
  103. builds leave the last successful model visible. Versioned `viewer.html`
  104. snapshots remain available; from v2,
  105. previous and current are preloaded for side-by-side, overlay, and
  106. Parts-tab change review (`--no-view` opts out)
  107. - **Spatial review comments** — pin human feedback to a surface and named part;
  108. comments persist locally for the agent's next revision
  109. - **`agentcad review list --status open`** — read shared review threads; humans
  110. and agents can reply, resolve, and reopen with an auditable history
  111. - **`agentcad parts view REF`** — hand off an isolated, focused, or grouped part review viewer
  112. ### Deliver
  113. - **`agentcad run ... --export stl,glb`** or **`agentcad export`** — write mesh
  114. deliverables and report validation for each written format
  115. - **`agentcad viewer [open|status|stop]`** — open the live project or manage its
  116. lightweight local service; `open` is the default
  117. - **`agentcad docs [section]`** — runtime-aware built-in documentation and worked examples
  118. ## Spatial review comments
  119. Live project viewers opened after a run or with `agentcad viewer open` use the existing service bound to
  120. `127.0.0.1`. Click **Comments** in the bottom-left viewer controls, then **+ New comment**
  121. (or press `C`). Click the model, confirm the inferred named part and
  122. revision scope, then choose **Send comment**. To hold feedback back, choose
  123. **Save draft** instead; unsent drafts can be edited, deleted, sent individually,
  124. or submitted together with **Send all drafts**.
  125. Comments are plain local JSON under the selected build root's `.agentcad/reviews`—there is no database or
  126. hosted account. On its next turn, an agent can discover and reply to the review:
  127. ```bash
  128. agentcad context
  129. agentcad review list --status open
  130. agentcad review comment --message "Could this rib be thinner?" --part support_rib --version current
  131. agentcad review reply C1 --message "Increased the clearance to 4 mm." --version current
  132. agentcad review resolve C1 --message "Implemented in the current revision." --version current
  133. ```
  134. Agents can also initiate an immediately-open thread on a named part. Add
  135. `--scope previous|both` to target comparison sides or `--point-mm X,Y,Z` to
  136. place its pin more precisely. Humans and agents can both reply, resolve, or
  137. reopen. Every action records its
  138. actor, timestamp, optional message, and associated revision. The older
  139. `mark-addressed` command remains available for compatibility.
  140. New builds wait while you write or send a comment or reply, then update the
  141. model once you send, save, or cancel. Current/Previous/Both follow the displayed
  142. revision sides; the original source version stays in the thread. If a part
  143. disappears, its comment remains in the list as unplaced. Hiding a part hides its
  144. comment pins in the 3D view; its threads stay in the sidebar. Showing the part
  145. restores its pins. Comments become
  146. available to the agent on its next turn; they do not start a new agent run.
  147. File viewers and temporary part/diff viewers remain read-only.
  148. To reposition the model, click **Pan** and drag, or hold **Shift** while dragging.
  149. Click **Pan** again to return to orbiting. **Fit view** centers the visible model
  150. and adjusts the zoom to fit it on screen.
  151. `--label` names the version; the JSON response returns the actual file under
  152. `outputs.step`. The older `--output LABEL` spelling remains a deprecated
  153. compatibility alias and never denotes a destination path.
  154. ### Keep one preview open
  155. After a normal `run` or `import`, give the human `project_viewer.url` from the
  156. JSON response. That URL tracks the latest completed viewer; `viewer` still
  157. points to this version's snapshot. An active page prevents another browser tab
  158. opening. Updates are checked about once a second (less often in background
  159. tabs). A new model replaces the previous one only after it loads successfully.
  160. `--no-view` never launches a browser or starts the viewer service. If it still
  161. generates viewer artifacts, an already-open project page can receive them.
  162. The core-only `--no-preview --no-diff --no-view` combination generates no viewer;
  163. the page retains the older model and reports that the new preview is unavailable.
  164. The service binds to `127.0.0.1` and serves only registered viewer snapshots at
  165. private project URLs. These are local bookmarks, not public share links. URLs
  166. survive service restarts and builds, but moving the project folder changes its
  167. identity. Viewer state lives in `~/.cache/agentcad/viewer` (override with
  168. `AGENTCAD_VIEWER_HOME` for isolated automation). Keep that directory to preserve
  169. the saved port and private URLs. If the port is occupied, AgentCAD reports the
  170. problem without silently changing the URL. Use `agentcad viewer status` and
  171. `agentcad viewer stop` for diagnostics; `agentcad viewer open` restarts it.
  172. An incompatible service protocol requires stopping the service with the
  173. installation that started it before reopening with the new installation.
  174. ## No boilerplate
  175. Scripts need zero imports. By default, build123d primitives, `show_object`, and agentcad edit helpers are pre-injected:
  176. ```python
  177. box = Box(10, 20, 5)
  178. show_object(box)
  179. ```
  180. For editor completion or explicit dependencies, the same AgentCAD-owned
  181. callables are available from the stable authoring module:
  182. ```python
  183. from agentcad.api import load_step, safe_cut, translate, show_object
  184. ```
  185. Import build123d primitives such as `Box` or `Vector` from `build123d` itself.
  186. Do not import STEP writers: `show_object` hands the result back to AgentCAD,
  187. which writes and tracks the canonical STEP file.
  188. In build123d scripts, capture each shape object once, including across
  189. `show_object()`, `show_assembly()`, and `show_compound()` calls and within
  190. assembly iterables. Capturing an assembly also captures all its descendants;
  191. do not capture a child separately or reuse it in another captured assembly.
  192. Repeated references fail with `error_kind: "duplicate_capture"`;
  193. changing the name or ID does not create another instance. For intentional
  194. instances, use `from copy import deepcopy`, then copy the shape before positioning
  195. and capturing it. See `agentcad docs parts` for an example.
  196. `agentcad init` records build123d as the project runtime. That keeps the
  197. script API, built-in docs, and subsequent runs on one clear default.
  198. ## CadQuery compatibility
  199. CadQuery remains supported for existing scripts and projects, but it is not
  200. the default authoring path and it is not installed by default. `pip install
  201. agentcad` ships build123d and the OpenCascade binding only. Enable the
  202. compatibility runtime with the optional extra:
  203. ```bash
  204. pip install "agentcad[cadquery]"
  205. agentcad daemon restart # if a daemon was already running
  206. ```
  207. The extra pulls in CadQuery and its own dependency chain (CasADi and the
  208. COIN-OR solver libraries it bundles), which carries a larger third-party
  209. license surface than the default install. Selecting the CadQuery runtime
  210. without the extra returns a structured error naming this command. Adopting
  211. agentcad inside a product or hosted service? See
  212. [Incorporating agentcad into your project](https://agentcad.dev/incorporating-agentcad).
  213. For a CadQuery project:
  214. ```bash
  215. agentcad init --name legacy-model --runtime cadquery
  216. agentcad docs quickstart --runtime cadquery
  217. agentcad run script.py --label first
  218. ```
  219. For a one-off CadQuery script inside a build123d project:
  220. ```bash
  221. agentcad docs preamble --runtime cadquery
  222. agentcad run legacy.py --label legacy --runtime cadquery
  223. ```
  224. Keep each script on one CAD API. Conflicting imports, including mixed imports,
  225. are rejected against the project runtime before daemon startup or version
  226. creation, with guidance for fixing the script or intentionally overriding the
  227. runtime. Unpinned projects still detect the runtime and reject ambiguous scripts.
  228. Run responses identify the selection in `runtime_source` (`command`, `project`,
  229. or `detection`, which includes the default fallback). Run
  230. `agentcad docs runtimes` for the complete dispatch contract.
  231. ## MCP integration
  232. For native tool integration with Claude Code, Cursor, or Windsurf:
  233. ```bash
  234. pip install agentcad[mcp]
  235. ```
  236. Extras combine: `pip install "agentcad[mcp,cadquery]"` installs both.
  237. Add to `.mcp.json`:
  238. ```json
  239. {"agentcad": {"command": "python", "args": ["-m", "agentcad.mcp"]}}
  240. ```
  241. For local STEP/STP/BREP files, call `import_cad` with `file`, `cwd`, and
  242. `initialize=True` to start a project. Imports prepare typed geometry IDs,
  243. measurements and a compact inspection report by default. No Inspect opt-in is
  244. required. The default response includes the import
  245. JSON and up to nine inspection PNG attachments. Use `report_images=False` for
  246. an image-free report or `report=False` to keep the bundle on disk. Image
  247. attachments are delivered to the MCP client; how they appear to the user depends
  248. on the client. Follow the
  249. returned IDs and paths with `inspect` and `run`. Local paths refer to the MCP
  250. execution filesystem. See `agentcad docs mcp` for controls.
  251. For retry-safe imports, pass `idempotency_key="supplier-001"` (CLI:
  252. `agentcad import supplier.step --init --idempotency-key supplier-001`). Preserve
  253. the key and identical source/options after a lost response: retries return the
  254. original version without importing or opening the viewer again. A different
  255. source, label, runtime request, validation profile, report/image limits or
  256. viewer/diff options conflicts with that key. Explicit checks JSON content and
  257. `--defer-checks` must also match. Use a new key for an intentional
  258. new import. Keys are scoped to the selected build directory; the original file
  259. must remain readable with identical bytes.
  260. A contended retry waits up to ten seconds, then reports `in_progress` if the
  261. original process still owns execution; retry the same request later. A stopped
  262. process releases ownership automatically. Retrying recovers a committed version
  263. and marks unfinished artifacts interrupted, or explicitly reports that execution
  264. stopped before commit. Optional analysis is not automatically resumed. Saved
  265. responses describe the original attempt; use `inspect` for fresh inspection.
  266. When retrying from another project subdirectory, returned actions target the
  267. selected baseline from that caller directory. Inline detail/media may be trimmed
  268. further to keep the final report within its text budget; saved analysis and IDs
  269. stay unchanged. If required references cannot fit, `inspection_delivery` reports
  270. the omission and provides a retrieval action with a larger allowance.
  271. Deleting `.agentcad/import_requests` deletes retry protection for those keys.
  272. Imports return a durable `inspection_job` and wait for inspection by
  273. default. Use `--inspection-background` or `--inspection-wait-timeout SECONDS`
  274. to return before analysis finishes. A wait deadline leaves the worker running;
  275. follow the returned status/wait/resume/cancel actions. These wait controls do
  276. not change import retry identity. See [inspection jobs](docs/inspection-jobs.md)
  277. for CLI/MCP examples, checkpoints and cancellation behavior.
  278. Import and `inspect --report` default to a 2,000-estimated-token limit for the
  279. complete JSON response, with bounded image samples and a full-response retrieval
  280. action. `--full-response` opts out of whole-response compaction. Agents can supply `--context-window`, `--response-budget`,
  281. `--response-text-budget`, `--image-budget` and a caller-estimated
  282. `--vision-profile` to bound complete inspection delivery. Limited responses
  283. include a full saved response and an executable retrieval action. See
  284. [inspection budgets](docs/inspection-budgets.md) for estimates, CLI/MCP parity
  285. and the distinction between whole-response limits and report preparation.
  286. Saved response snapshots default to 30 days and 64 MiB per build root, with
  287. configurable limits and explicit expiry/eviction responses. Cleanup preserves
  288. geometry, issued IDs, import retry records and resumable jobs.
  289. Analysis, image sheets and completed job results also have shared
  290. [cache limits](docs/inspection-cache.md), defaulting to 30 days/512 MiB per build
  291. root. Resumable checkpoints remain protected; a full quota stops enrichment
  292. with a resume action while preserving the imported model.
  293. To retrieve the complete inventory without opening server-side files, call
  294. `inspect(artifact="manifest", file=..., cwd=...)` (CLI: `inspect FILE --artifact
  295. manifest`). Saved report/image paths and `artifact="baseline"` use the same
  296. bounded byte transfer, with checksums and copyable continuations. See
  297. [inspection artifacts](docs/inspection-artifacts.md). This works through the
  298. existing MCP tool connection; the separate `cloud_*` API is unchanged.
  299. ### Cloud projects
  300. For setup and a complete agent build-and-edit loop, see the
  301. [cloud workflow guide](docs/cloud-workflow.md).
  302. If your host supplies a hosted MCP URL, a supported agent can connect directly
  303. with the issued key pair, without installing AgentCAD locally. Call `cloud_help`
  304. for remote tools and explicit content uploads. The guide covers URL setup and
  305. retrieving private evidence through MCP.
  306. If your host has issued an AgentCAD API address and key pair, the same local
  307. MCP server also offers `cloud_run`, `cloud_import`, `cloud_measure`,
  308. `cloud_inspect`, `cloud_check_spec`, `cloud_render`, and `cloud_diff`. Project,
  309. job, and download tools support that workflow. Set `AGENTCAD_API_URL`,
  310. `AGENTCAD_KEY_ID`, and `AGENTCAD_KEY_SECRET` in the server process environment.
  311. Keep credentials in a password manager, outside project files and agent prompts.
  312. For the local cloud CLI, install
  313. `pip install 'agentcad[cloud]==0.8.0'` (also included in `agentcad[mcp]`),
  314. as shown in the guide above:
  315. ```bash
  316. agentcad cloud projects create --name Bracket
  317. agentcad cloud run PROJECT examples/cloud/bracket.py --label first
  318. agentcad cloud job PROJECT JOB --wait
  319. agentcad cloud measure PROJECT 1
  320. agentcad cloud inspect PROJECT 1 --summary
  321. agentcad cloud check-spec PROJECT 1 examples/cloud/spec.json
  322. agentcad cloud render PROJECT 1
  323. agentcad cloud download PROJECT iso.png --job RENDER_JOB --kind render
  324. ```
  325. Replace `PROJECT` and `JOB` with the returned IDs. Submissions return immediately;
  326. poll the same job until it completes. Keep its `idempotency_key` for an identical
  327. retry after a lost response. Upload an edited script with a new key and label to
  328. create another version, then compare with `agentcad cloud diff PROJECT 1 2`.
  329. Only the named script, STEP, or spec file is uploaded. No local project is needed.
  330. With matching client/host export support, `agentcad cloud export PROJECT VERSION
  331. --format stl,obj,glb` exports a saved whole model without rebuilding it. Poll
  332. that export job, read its per-format mesh validation, and download with
  333. `--kind stl`, `obj`, or `glb`. See the [export workflow](docs/cloud-workflow.md#export-a-saved-model)
  334. for availability, units, private retrieval, and limits.
  335. Downloads require an explicit destination and preserve existing files by default.
  336. Job completion, geometry validity, spec verdicts, and human approval remain
  337. separate. Run `agentcad docs cloud` for all commands, limits, and MCP tool names.
  338. Cloud access uses your host's installed CAD engine and credentials; it is not a
  339. hosted remote MCP endpoint or a general remote mode for local commands.
  340. ## Requirements
  341. - Python 3.10–3.12 (OpenCascade bindings do not support 3.13+)
  342. ## License
  343. Apache License 2.0. See [LICENSE](LICENSE).
  344. ## Feedback
  345. If your agent struggles, run `agentcad feedback "what happened"` to capture a friction log.
  346. For a detailed session report, create and fill in a guided feedback template:
  347. ```bash
  348. agentcad feedback --template --output feedback.md
  349. # Fill in the task, what happened, evidence, and any workaround.
  350. agentcad feedback --file feedback.md
  351. ```
  352. Creating a template sends nothing. Reports include recent session context;
  353. review what you send, or add `--local-only` to keep it local. Partial reports,
  354. positive feedback, and ideas are welcome. See `agentcad docs feedback` for
  355. file/stdin input and upload receipts.

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.