Compatibility and testing

Which platforms, slicers, agents and programs Form Studio works with, and what the continuous integration exercises on every change.

Status labels

Label Meaning
CI Exercised automatically by a GitHub Actions workflow, with the real program where named
Supported Implemented and covered by unit tests against fixtures; not run by CI with the real program

Platforms

Platform Engine tests Packaged app starts Interface journeys Real slices
macOS, Apple Silicon CI CI Supported Supported
macOS, Intel CI CI Supported Supported
Windows 10 and 11, x64 CI CI CI CI
Linux x64 (glibc 2.35+) CI CI CI CI

Slicers

Slicer Version in CI Linux Windows macOS Notes
Bambu Studio 2.9.1 CI CI Supported Presets from the slicer's selection or pinned in Settings
OrcaSlicer 2.5.1 CI CI Supported Packed vendor profiles (.opc) read directly
PrusaSlicer Ubuntu package (Linux), Chocolatey (Windows) CI CI Supported Bundled vendor profiles used without opening the slicer
SuperSlicer 2.5.59 CI CI Supported
QIDISlicer 1.2.5 CI CI Supported
ElegooSlicer 1.5.3.5 CI CI Supported OrcaSlicer family
Snapmaker Orca 2.4.1 CI CI Supported OrcaSlicer family

Every printer set up in one of these slicers can be chosen, per design or as the default. Time and filament are the slicer's own estimate and are labelled simulated.

Every night, every recipe (built-in and community) is sliced by every slicer above on Linux, and the result of each recipe and slicer pair (built, sliced, checks passed, print time, filament, layers) is published as the compatibility matrix. The pipeline's timings per recipe and the app's start time are tracked on every change to dev and drawn as benchmarks.

Agents

Agent Chat inside the app MCP from the agent's own window Notes
Claude Code Supported Supported Sign-in from the app
Codex Supported Supported
Gemini CLI Supported Supported Sign in, or paste an API key in Settings › Agents
Cursor CLI Supported Supported
OpenCode Supported Supported
Qwen Code Supported Supported
GitHub Copilot CLI Supported Supported
Model APIs: Anthropic, OpenAI, Gemini, OpenRouter, Mistral, xAI, DeepSeek, Groq Supported Found from your key; Form Studio runs the conversation and its tools
Local models: Ollama, LM Studio, llama.cpp Supported Found from their running servers
Any provider added by its address Supported OpenAI-, Anthropic- or Gemini-compatible, or Ollama; format and models detected
Claude Desktop Supported
Any MCP client over stdio Supported { "mcpServers": { "form": { "command": "form-mcp" } } }

The MCP server itself is tested in CI over real stdio (tests/test_mcp.py): tools are listed and called, renders come back as images, a stale expect_head is refused, and no tool can approve a release.

Other programs

Program Used for Status
build123d on OpenCascade Every solid, STEP export and B-rep check CI
FreeCAD 1.x Open the STEP, re-import check, parameter macro Supported
OpenSCAD .scad designs with Customizer parameters Supported
Blender Photoreal renders; 3D-Print Toolbox checks when installed Supported
STEP and STL files Reference parts to design around CI

Continuous integration

Workflow Runs On
CI The test suite; checks that the MCP reference matches the server Linux, macOS; every push to main and every pull request
Desktop app Engine tests, the app bundle build, the packaged app launch macOS arm64 and x64, Windows, Linux
Interface tests The interface journeys with a real PrusaSlicer Windows, Linux
Real slicers A real slice with each slicer's pinned official build, verified by SHA-256 Windows, Linux
Compatibility matrix Every recipe through every slicer CI installs Linux, nightly
Benchmarks Pipeline time per recipe and app start time Linux, every change to dev
Release smoke test The latest published release, downloaded and checked against SHA256SUMS, then opened and used All four platforms, nightly
Recipe check A recipe submitted in an issue, built, checked and sliced, answered with pictures Linux, on the issue
PR preview Before and after pictures of the recipes and interface a pull request changes Linux, on the pull request
CodeQL Static analysis of the engine, the app, the website's JavaScript and the workflows Every push and pull request
Links Links in the docs and the website Pull requests; every link weekly
Website Docs build, recipe index, link and page-weight check Linux

Running the tests yourself

python -m unittest discover -s tests          # the suite; slicing tests run when a slicer is installed
FORM_UI_TESTS=1 python -m unittest tests.test_ui_journeys   # the interface journeys in the real window
python scripts/acceptance.py                  # end to end: from the first revision to the release, recovery, rebuild latency

What results mean

Form Studio labels every number with how it is known. A geometric check is computed on the kernel's solid. A slice is simulated. Fit yield and tip-over are models with stated assumptions. Only a measurement from a real print, entered by you, is physical; coupon measurements calibrate the fit model for your printer and filament.