A pure Python library for offline generation of Max/MSP patcher files (.maxpat, .maxhelp, .rbnopat).
If you are looking for Python 3 externals for Max/MSP, check out the py-js project.
pip install py2maxFor the browser-based live editor and remote REPL, install the companion
py2max-server package:
pip install py2max-serverFor development:
git clone https://github.com/shakfu/py2max.git
cd py2max
uv sync
source .venv/bin/activatefrom py2max import Patcher
p = Patcher('my-synth.maxpat')
osc = p.add('cycle~ 440')
gain = p.add('gain~')
dac = p.add('ezdac~')
p.link(osc, gain)
p.link(gain, dac)
p.save()That's it! Open my-synth.maxpat in Max to see your patch.
- Offline Patch Generation - Create Max patches programmatically without Max running
- Round-trip Conversion - Load, modify, and save existing
.maxpatfiles - Max for Live (.amxd) - Read/write binary
.amxddevice files with presentation-mode helpers - Universal Object Support - Works with any Max/MSP/Jitter object
- Fully typed - Passes
mypy --strict; no runtime dependencies - High Test Coverage - 420+ tests ensure reliability
Generate Max for Live devices directly. Patcher.save() / Patcher.from_file()
auto-detect the .amxd extension and read/write the binary device format,
byte-for-byte compatible with Max-exported devices.
from py2max import Patcher
# device_type: "audio_effect" (default), "instrument", or "midi_effect"
p = Patcher('gain.amxd', device_type='audio_effect')
p.enable_presentation(devicewidth=120) # render Ableton's device strip
plugin = p.add_textbox('plugin~') # audio in from Live
gain = p.add('live.gain~', maxclass='live.gain~')
plugout = p.add_textbox('plugout~') # audio back to Live
gain.add_to_presentation([20, 20, 60, 136]) # show the fader in the device
p.add_line(plugin, gain, outlet=0, inlet=0)
p.add_line(gain, plugout, outlet=0, inlet=0)
p.save() # writes a binary .amxdHelpers: Patcher.enable_presentation(devicewidth=...),
Box.add_to_presentation([x, y, w, h]) (rejects M4L infrastructure objects and
rounds fractional coordinates), and Patcher.enforce_integer_coords(). M4L
binary helpers live in py2max.m4l.
Real-time browser-based patch editing with bidirectional sync lives in the
companion py2max-server package, so
the core library stays small and offline:
pip install py2max-server
py2max-server serve my-patch.maxpat
# Opens browser at http://localhost:8000Features:
- Drag objects, draw connections visually
- Three layout engines: WebCola, ELK, and Dagre
- Auto-save with debouncing
- Navigate into subpatchers
- REPL mode for Python interaction
Generate high-quality SVG previews without Max:
py2max preview my-patch.maxpat --openp = Patcher('synth.maxpat')
# ... add objects ...
p.to_svg('synth.svg', title="My Synth", show_ports=True)Five built-in layout strategies:
| Layout | Description |
|---|---|
grid |
Connection-aware clustering with configurable flow |
flow |
Signal flow-based hierarchical positioning |
columnar |
Controls -> Generators -> Processors -> Outputs |
matrix |
Signal chains in columns, categories in rows |
horizontal/vertical |
Simple grid layouts |
p = Patcher('patch.maxpat', layout='flow', flow_direction='vertical')
# Add objects and connections...
p.optimize_layout() # Arrange based on signal flow
p.save()Access documentation for 1157 Max objects:
p = Patcher('demo.maxpat')
cycle = p.add('cycle~ 440')
print(cycle.help()) # Full documentation
print(f"Inlets: {cycle.get_inlet_count()}")
print(f"Outlets: {cycle.get_outlet_count()}")Optional validation catches wiring errors:
p = Patcher('patch.maxpat', validate_connections=True)
osc = p.add('cycle~ 440')
gain = p.add('gain~')
p.link(osc, gain) # Valid
p.link(osc, gain, outlet=5) # Raises InvalidConnectionErrorHuman-readable object IDs for easier debugging:
p = Patcher('patch.maxpat', semantic_ids=True)
osc1 = p.add('cycle~ 440') # ID: 'cycle_1'
osc2 = p.add('cycle~ 220') # ID: 'cycle_2'
gain = p.add('gain~') # ID: 'gain_1'
# Find by semantic ID
osc = p.find_by_id('cycle_1')Query Max object metadata efficiently:
from py2max import MaxRefDB
db = MaxRefDB() # Auto-cached on first use
print(len(db)) # 1157 objects
if 'cycle~' in db:
info = db['cycle~']
print(info['digest'])
# Search and filter
results = db.search('filter')
msp_objects = db.by_category('MSP')from py2max import Patcher
p = Patcher('my-patch.maxpat')
osc = p.add('cycle~ 440')
gain = p.add('gain~')
dac = p.add('ezdac~')
p.link(osc, gain)
p.link(gain, dac)
p.link(gain, dac, inlet=1) # Stereo
p.save()p = Patcher.from_file('existing.maxpat')
# Find and modify objects
for box in p.find_by_text('cycle~'):
print(f"Found oscillator: {box.id}")
p.save_as('modified.maxpat')p = Patcher('main.maxpat')
sbox = p.add_subpatcher('p mysub')
sp = sbox.subpatcher
# Build the subpatcher
inlet = sp.add('inlet')
gain = sp.add('gain~')
outlet = sp.add('outlet')
sp.link(inlet, gain)
sp.link(gain, outlet)
# Connect in main patcher
osc = p.add('cycle~ 440')
dac = p.add('ezdac~')
p.link(osc, sbox)
p.link(sbox, dac)
p.save()p = Patcher.from_file('complex-patch.maxpat')
# Find by ID
obj = p.find_by_id('obj-5')
# Find by text content
oscillators = p.find_by_text('cycle~')
# Find by object type
messages = p.find_by_type('message')# Create new patch from template
py2max new demo.maxpat --template stereo
# Show patch info
py2max info demo.maxpat
# Generate SVG preview
py2max preview demo.maxpat --open
# Optimize layout
py2max optimize demo.maxpat --layout flow
# Validate connections
py2max validate demo.maxpatProvided by the separate py2max-server
package (pip install py2max-server):
# Start server with browser editing
py2max-server serve my-patch.maxpat
# With REPL in same terminal
py2max-server serve my-patch.maxpat --repl# Show cache status
py2max db cache location
# Create category-specific database
py2max db create msp.db --category msp
# Search objects
py2max db search maxref.db "oscillator" -v
# Query specific object
py2max db query maxref.db cycle~ --json# Convert .maxpat to Python code
py2max convert maxpat-to-python patch.maxpat output.py
# Lookup object documentation
py2max maxref cycle~ --json- Scripted patch generation - Automate repetitive patch creation
- Batch processing - Modify multiple
.maxpatfiles programmatically - Parametric patches - Generate variations from configuration files
- Test generation - Create
.maxhelpfiles during external development - Container population - Prepopulate
coll,dict,tableobjects with data - Generative patching - Algorithmic patch creation
- CI/CD integration - SVG previews for documentation and version control
make test # Run all tests
make typecheck # Type checking with mypy
make lint # Linting with ruff
make docs # Build documentationThe .maxpat JSON format maps directly to three Python classes:
Patcher- The patch container with boxes and patchlinesBox- Individual Max objectsPatchline- Connections between boxes
All classes are extendable via **kwargs, allowing any Max object configuration. The add_textbox() method handles most objects, with specialized methods (add_subpatcher(), add_coll(), etc.) for objects requiring extra configuration.
- Max doesn't refresh from file when open - close and reopen to see changes, or use
py2max-server serve(from the separatepy2max-serverpackage) for live editing - For tilde variants, use the
_tildesuffix:p.add_gen()vsp.add_gen_tilde() - API docs in progress - see
CLAUDE.mdfor comprehensive usage
The tests/examples/ directory contains working, tested
examples organized by topic (see its README):
quickstart/basic_patch.py- Simple oscillator patchtutorial/signal_processing_chain.py- Complex audio processing chaintutorial/generative_music.py- Generative music system with patternslayout/grid_layout_examples.py- Grid layout with clusteringadvanced/data_containers.py- Tables, collections, and dictionariesapi/patcher_api_examples.py- Patcher API reference examples
External usage:
- faust2rnbo - Generate Max patchers for RNBO
We welcome contributions! See CONTRIBUTING.md for guidelines.
git clone https://github.com/shakfu/py2max.git
cd py2max
uv sync
source .venv/bin/activate
make test # Verify setupMIT License. See LICENSE for details.
- HOLA algorithm: Kieffer, Dwyer, Marriott, Wybrow (IEEE 2016)
- NetworkX: Hagberg, Schult, Swart (SciPy 2008)
- Graph drawing techniques: Gansner, Koutsofios, North, Vo (IEEE 1993)