Slicing Guide

H2D Pro uses BAMBU_MODEL=h2dpro and requires the selected slicer installation’s exact Bambu Lab H2D Pro <nozzle> nozzle preset. Offline tests cover model selection, preset preparation and mocked external-spool dispatch; physical H2D Pro printing is not verified. Do not substitute h2d.

TL;DR

For offline prepared-file evidence, call inspect_print_file with file_path, explicit bambu_model, and optional zero-based plate_index, nozzle_diameters and bed_type. It inspects a private read-only snapshot using the existing safety engine and returns a versioned SHA/byte-length/plate receipt, declarations and checked temperature peaks. It neither contacts a printer nor slices input. evidence: "file-only" and dispatch_authorized: false mean the result cannot replace fresh printer preflight or human authorization; actual printing independently checks its final dispatched snapshot.

Printing includes a mandatory preflight of the final selected G-code and fresh MQTT reports. The requested model, file model, printer identity, and reported nozzle diameters must agree. Missing or contradictory safety information stops the operation. A nozzle reported abnormal or unknown blocks printing. A nozzle with an identity chip whose information the printer marks unreliable must be refreshed on the printer first. Stock nozzles without an identity chip (serial N/A, such as H2D HS01 hotends) always carry that unreliable flag. Their reported diameter and type are configured settings, so they are still compared with the job and named in the print prompt for a person to confirm. Positive heating commands must fit independent hardware and material ceilings, including later commands and waiting-temperature targets. These ceilings allow supported purge routines; they are not recommended print temperatures.

Print, heating and resume start without a prompt; clearing hardware errors and resuming a file that could not be re-read still ask a person. The client's own Print button is the yes, and the printer runs its own checks; a printer reporting FINISH does not add a separate plate confirmation. Clearing hardware errors returns the reported codes, and the next print then asks a person to confirm those causes are resolved. That prompt shows the verified model, nozzle diameters, declared materials, temperature peaks, and file hash. Clients without elicitation support cannot bypass hardware-error confirmation. A declined or cancelled prompt is reported as declined, not as missing client support: some clients decline prompts automatically, for example in a non-interactive permission mode. A prompt waits ten minutes for an answer by default (BAMBU_CONFIRMATION_TIMEOUT_MS); an unanswered prompt is reported as a timeout and nothing is sent to the printer. Stop and heater-off remain available without confirmation.

PLA uses a 260°C normal/manual policy ceiling. The inspector permits one bounded, exact X1E startup purge sequence up to 290°C, then requires normal targets before deposition. This is command-form validation, not a simulation of firmware timing. Explicit laser/cutting job metadata and recognized laser-enabling commands are refused; this server supports FFF printing.

Archives with duplicate, case-colliding, hidden, or inconsistent entries are rejected. The dispatched plate path and checksum come from the inspected entry. Inspection accepts files up to 256 MiB and archives up to 512 MiB uncompressed; ZIP64 is unsupported. Printable upload-only operations inspect every plate and verify live model/nozzles, use unique remote names, and refuse existing remote destinations. Upload inspection does not authorize a later print or establish its AMS mapping.

Pre-sliced dual-nozzle jobs retain their complete diameter list. Pass nozzle_diameters: [0.4, 0.6] to explicitly require that configuration, or omit nozzle arguments to compare the file's declared diameters directly with live telemetry. A scalar nozzle_diameter requires that size throughout the job. Mixed-diameter projects require GUI slicing before printing; the CLI preset path supports a uniform diameter.

Use a sliced file carrying its machine, nozzle, filament, and bed metadata. Declare non-RFID materials in the slicer and supply the complete physical AMS mapping or select external-spool printing. Available printer-reported materials are compared with that declaration. Reports cannot prove the contents of a manually labelled spool, the physical nozzle behind a manual setting, or that the build plate is clear. The checks validate declared object bounds and supported thermal syntax; they do not simulate arbitrary motion or authenticate a file's metadata.

Legacy P1/X1/A1-mini .gcode.3mf transport cannot send an AMS mapping or select a plate. It accepts a single external-spool-only plate; export a .3mf project for checked AMS mappings. Remote .gcode starts download and inspect the file, then print a uniquely named checked copy. The optional BambuNetwork wrapper also requires local MQTT telemetry even for cloud dispatch. Its raw mutation and parameter-override routes are restricted.

Resume accepts an inspected job started by this server through either direct printing or the BambuNetwork wrapper, after fresh paused-job identity and configuration checks. Bridge submissions use a unique task name as well as a unique remote filename; project_name (or the legacy task_name label) remains the project label.

These preflight changes have automated regression coverage with intercepted printer transports. Earlier live-print evidence below predates this additional gate; no new physical print is claimed by those tests.

Use case	Path	Status
Single-colour BambuStudio CLI slicing	MCP slices via CLI with automatic BBL profile resolution	✅ Works (verified H2S, H2D, X1C, P1S on 02.06.01.55). H2C requires Bambu Studio 2.4.0+ and BAMBU_MODEL=h2c.
FULU/Orca CLI preparation (1.1.11+)	Same exact-machine gate and BBL dependency resolution	Regression-tested preparation and failure handling; live slicer/version validation remains separate.
Multi-color BambuStudio CLI slicing	MCP prepares every filament slot's colour and a fallback tower position	Contributor-verified on Windows BambuStudio 02.08.02.60, including four-colour X2D slicing. See version limits below.
X2D status and slicing	BAMBU_MODEL=x2d with an installed X2D preset	Available. Native printing additionally requires macOS or x86-64 Linux and the optional helper; legacy FTPS remains unsupported.
Pre-sliced .gcode.3mf → printer	MCP print_3mf	✅ Works (verified live on Kingpin H2D)
Other slicing combinations	Pre-slice in the GUI, then use print_3mf for a supported printer	Inspect the preview and use the printer's supported transport.

Choose GUI export or CLI slicing. FULU export plus the direct LAN print path does not require Bambu Studio, Bambu Connect, or Bambu Cloud. See FULU setup for that workflow and the separate optional BambuNetwork bridge.

Path A — pre-slice in FULU OrcaSlicer-bambulab, OrcaSlicer, or Bambu Studio:

Mesh ──► chosen slicer GUI ──► sliced .gcode.3mf ──► MCP print_3mf
         slice + preview + export

Path B — let the MCP slice via BambuStudio, FULU, or Orca CLI (BBL printers only):

STL/3MF ──► MCP slice_stl / print_3mf ──► (auto-flatten profiles) ──► chosen CLI ──► sliced .gcode.3mf

Path B works because the MCP now flattens BBL profile inheritance before calling the CLI — a workaround for several upstream bugs in BambuStudio's CLI mode (issues #9636 and #9968). Verified on H2S, H2D, X1C, and P1S with stock BBL profiles for single-colour slicing. The newer multi-colour evidence and version limits are described below. H2C is accepted as BAMBU_MODEL=h2c; use Bambu Studio 2.4.0 or newer for the H2C printer preset and do not substitute h2d.

BBL profile resolution now runs automatically for CLI slicing. BAMBU_CLI_FLATTEN is no longer required and cannot disable resolution. Missing parents, missing or malformed includes, cycles, and unresolved filament slots stop the slice before the CLI runs; the MCP does not fall back to partial profiles. Standalone custom process and filament configs remain usable, and custom BBL-derived profiles retain their settings on top of resolved parents. For BambuStudio, FULU, and Orca CLI tools, slicer_profile supplies process settings; the selected model's bundled machine preset must still be available. It is not a replacement machine configuration or a way to bypass model validation. Machine settings carried by a template's project_settings, a custom process file, or a custom filament file (for example another printer's start G-code, printer_model, or nozzle values) are dropped and logged before the CLI runs; the selected machine preset keeps them. Pre-sliced 3MF printing does not require running this CLI profile preparation.

Profile discovery follows the active executable: macOS app bundles, Windows resources/profiles, and Linux share/BambuStudio/profiles / share/OrcaSlicer/profiles layouts are recognized. For AppImages or other layouts whose profiles are not accessible beside the executable, set BAMBU_PROFILES_ROOT to the matching installation's directory containing BBL. An unavailable tree produces an error rather than using another installation's settings.

Orca cache-based installations

Local Orca builds that ship a version-2 BBL.opc instead of machine/process/filament JSON can use the selected installation directly. Set SLICER_TYPE=orcaslicer and SLICER_PATH to that build. The package must also contain BBL/cli_config.json: earlier cache-only packages without it still stop before slicing. Unknown cache versions, corrupt data and unresolved model/nozzle or profile dependencies fail explicitly.

The MCP validates and expands that installation's cache into its private temporary directory, then uses a separate private Orca data directory for generated User presets. It preserves custom settings and passes positional colours, bed selection and saved tower coordinates through Orca's project-level CLI options. Native extrusion mode stays enabled for these builds, allowing multi-colour wipe towers. Your installed profiles and running GUI are not changed. The remote orca-slicer-api route still requires its matching JSON profile mirror.

Verified software: the October 23 nightly, macOS universal asset 669077524, SHA-256 b7d0cc57d0ff5bc0663d48676939aa4ade6573301d25220d3476c3fe8f97406e. Real isolated P1S 0.4 single- and two-colour slices preserve custom temperatures, ordered colours and tower coordinates, and their exact output passes independent file/plate inspection. Cache/profile tests also cover A1 mini, H2D, H2C, X2D and P1S 0.2. Those tests do not establish physical printing or Linux/Windows slicer behavior. Stable/alpha packages are separate artifacts; this is not a claim that every Orca version works.

The OrcaSlicer 2.5.0 stable macOS universal release, source 40ecc66f78ed72dc368e0c1ebe40d77f00cc8a19, asset 719942288, SHA-256 061e0c6ff104c37655c83190725ac2420960f23c39b56aae9b546e6c542995d5, is also checked with real isolated CLI slices: P1S 0.4 single/two-colour, P1S 0.6 single-colour, and H2C/X2D 0.4 single-colour. Their original output passes file inspection; P1S two-colour output retains slot order, explicit colours, custom process settings and saved tower coordinates. For dual nozzles, Orca may write one shared nozzle_volume_type; inspection accepts that only with complete, matching physical extruder selections in the sliced plate header. See Orca's vector lookup and variant selection. These checks establish macOS slicing and inspection, not physical printing, dual-nozzle multi-material operation, or Linux/Windows CLI behavior. A2L remains unsupported by the print-safety model gate even though Orca can slice its preset.

Multi-colour CLI support and version limits

Bambu Studio 02.09.01.70 emits M104 M S<temperature> when the temperature changes after the first layer. The inspector accepts that marker while retaining independent heater and material limits. Isolated macOS P1S 0.4 single- and two-colour output from the official 2.9.1 release is checked against the exact generated files. Additional offline 0.4 mm, single-PLA cube slices for A1, A1 mini, P1P, P1S, P2S, X1C, X1E, H2S, H2D, H2D Pro, H2C and X2D passed the current file inspector using that exact release archive. These are slicer/file checks, not multi-material validation for those other models or physical printing. H3C (nine nozzles, serial prefix 32X) remains unsupported and fails the model gate.

The optional Linux X2D native helper pins networking plug-in 02.09.01.54 from Studio 2.9.1. Its PrintParams source layout, archive and required library checksums were verified. Linux plug-in loading and physical X2D dispatch were not tested for this version; rerun bambu-printer-mcp-native-setup after updating the package.

The CLI can crash when a project uses a filament after the first but its loaded profiles leave filament_colour at a single-entry default. The MCP now supplies one colour per slot. Explicit filament_colours values take priority, then the input 3MF's colours, each profile's own colour, and finally the BambuStudio default. Custom filament settings remain intact.

For multi-nozzle printers, an unset tower position is placed within the shared nozzle area. Saved project positions and explicit process positions are preserved. This does not guarantee a collision-free layout: inspect the slicer preview for the actual model and tower geometry.

Sebastian's contribution includes real Windows BambuStudio 02.08.02.60 bisection and a successful four-colour X2D slice. This is slicing evidence, not a physical-print test.

Older H2D failures remain separate evidence: 02.06.00.51 crashed during slicer setup, and 02.06.01.55 reported No valid nozzle found / code -100 on an exported multi-colour project. These were filed in BambuStudio#10408. The colour fix does not establish that every older-version or multi-material failure is resolved. Use a GUI-sliced project when a CLI combination fails.

X2D has its own preset, status identification, and slicing support. On macOS and x86-64 Linux, print_3mf uses the optional native helper and Bambu's networking plug-in, after the shared print safety checks. The job goes to a ready SD card or USB stick, otherwise to internal storage (eMMC). See native setup for macOS and Linux. Other operating systems, and Linux without the helper, reject native print requests before slicing or connecting. Legacy FTPS/remote-file starts remain unsupported; do not choose H2D as a substitute model. Existing H2S/H2D print routes remain unchanged.

Why Path A is still recommended

GUI slicing lets you inspect supports, colours, tool changes, and tower placement before printing. It is also the fallback when a CLI build rejects flags, cannot resolve the selected profiles, or crashes.

Bambu-compatible CLI preparation supports bundled BBL presets and standalone custom process/filament overrides; custom BBL-derived settings are retained after dependency resolution. The selected model's bundled machine preset is still required. FULU/Orca and their aliases use the same machine-preset gate from 1.1.11. See FULU/Orca CLI setup and validation limits.

A failed inspection or auto-slice stops before upload. The server never deliberately sends the original unsliced project as a fallback.

Path B mechanics (CLI auto-flatten)

Before BambuStudio, FULU, or Orca CLI slicing, the MCP:

1. Requires the exact model/nozzle machine preset in the selected installation's BBL tree, validates its identity, and reads each leaf profile JSON.
2. Resolves inherits and include recursively: inherited settings first, include templates in order, then the profile's own keys. Cycles and unresolved references are errors, including within templates.
3. Sets from: "User", inherits: <leaf machine name>, and printer_settings_id / print_settings_id / filament_settings_id so the CLI's compatibility check passes.
4. Derives nozzle settings from the selected machine profile and nozzle configuration. nozzle_volume_type (standard or high_flow) sets the machine's nozzle_volume_type, which selects each profile's Standard or High Flow values; a flow type the preset does not declare in extruder_variant_list is refused. Use the exact model preset; a successful slice for another model does not validate the target printer.
5. Auto-extends compatible_printers to include the chosen machine when the user picked a non-default printer/process combo.
6. Removes every key the resolved machine preset defines, other than profile metadata such as name, inherits, and compatible_printers, from the process and filament configs, because the CLI applies every key in every loaded file. For legacy JSON-based OrcaSlicer, absolute-extrusion normalization (use_relative_e_distances = 0, no per-layer G92 E0) is applied to the machine config. Cache-based Orca builds preserve their native extrusion mode and use the private preset-loading route described above.
7. Writes flattened temp configs and passes those paths to --load-settings / --load-filaments.

For project 3MF input, a single filament override is repeated across all declared project slots. An explicit list preserves order and duplicate paths and must provide one profile per slot. This prevents later slots from retaining a foreign printer's filament settings; it does not change object assignments or AMS tray mapping. A geometry-only 3MF without project settings has no embedded slot list to replace.

Implementation: src/slicer/profile-flatten.ts. Optional installed-slicer smoke test: node scripts/test-cli-slice.mjs --model h2s (substitute the supported target model). This slices a fixture; it does not print.

Remote Orca slicing

Set SLICER_TYPE=orca-slicer-api to use a separately hosted AFKFelix/orca-slicer-api service. The MCP host does not need an Orca executable or SLICER_PATH. It still needs a read-only copy or mount of the matching remote Orca installation'sresources/profiles tree, including BBL/cli_config.json and the exact model/nozzle machine preset:

SLICER_TYPE=orca-slicer-api
SLICER_API_URL=http://orcaslicer:3000
BAMBU_PROFILES_ROOT=/mnt/orca/resources/profiles
SLICER_API_MODE=sync
SLICER_API_TIMEOUT_MS=600000

SLICER_API_URL is trusted server configuration, never a tool argument. A proxy path prefix is supported. Use HTTPS when the connection needs transport protection. The URL cannot contain credentials, a query, or a fragment. For reverse-proxy authentication, set SLICER_API_TOKEN; the default header is Authorization: Bearer <token>. Set SLICER_API_TOKEN_HEADER=X-API-Key to send the raw token through a custom header. Redirects are rejected, returned job URLs are ignored, and remote errors do not echo response bodies or authentication values.

The backend targets the upstream API contract at a9a28f0a7b0c6f51b9be158509e9e9677532b868. sync submits multipart data to /slice. async submits once to /slice-async, polls /slice-async/{requestId}, downloads its /result, and deletes the finished job. The timeout covers submission, polling and downloading (default ten minutes; accepted range 1–3600000 ms). Timeout or cancellation does not cancel an active upstream job: the API has no active-job cancellation endpoint. The client reports that work may continue and never retries submission automatically.

Before HTTP submission, the same local Orca preparation resolves inherits and include recursively, validates the exact bundled machine and its CLI config, preserves custom process/filament overrides, and prepares colour and nozzle settings. The client uploads those complete machine, process and filament JSON files with resolveProfileInheritance=false; their retained leaf identity metadata serves the CLI compatibility check. The service cannot substitute its default profiles for missing local machine provenance.

The supported subset is STL or 3MF input with exactly one filament slot and one positive, one-based slice_plate (default 1). Uploaded profiles preserve custom settings and a single explicit/project/profile filament colour. Bed type, nozzle type and nozzle flow settings use the existing validated profile overlays. arrange=true and orient=true are supported. Explicit false for either is rejected because the pinned upstream multipart implementation treats nonempty strings as true. Requests for all plates (slice_plate=0), multiple filament slots, object/filament-ID remapping, repetitions, transforms, and CLI-only flags such as ensure_on_bed, uptodate, min_save, skip_modified_gcodes, enable_timelapse, and allow_mix_temp are rejected rather than dropped. Use a local backend when those controls are required.

Downloads are capped at 100 MB, saved under a unique private output name, and independently inspected as a sliced 3MF before being returned. The selected plate must retain its original Metadata/plate_N.gcode identity; renumbered results are rejected. The file must match the prepared model, nozzle diameters and material, and its commands must pass the existing independent safety inspection. Printing still performs its own fresh live-printer preflight. HTTP fixture tests verify the transport and failure boundaries; they do not establish a real Orca slice or physical print.

Why the sliced file matters

The slicer supplies model-specific start G-code, filament declarations, and plate metadata. Renaming a geometry-only .3mf to .gcode.3mf does not create those contents. An incorrectly prepared project may fail AMS validation, be rejected by firmware, or stop after heating.

Use an exported sliced project, or let the configured CLI finish successfully before upload. A .gcode.md5 checksum entry is not printable G-code.

The right input file

After slicing in your chosen GUI, export the sliced plate (in Bambu Studio, File → Export → Export plate sliced file, or "Export all sliced files"). Menu labels vary in FULU/Orca builds. A typical .gcode.3mf contains the following; the required printable content is a Metadata/plate_<n>.gcode entry, while the available metadata depends on the exporter:

Metadata/
  plate_1.gcode               ← the actual machine instructions
  plate_1.json                ← { "filament_ids": [...], ... }
  slice_info.config           ← <filament id="..."> declarations
  filament_sequence.json      ← per-plate filament order

If no Metadata/plate_<n>.gcode exists, print_3mf attempts auto-slicing. If the prepared output is still not printable, it stops with:

3MF does not contain any Metadata/plate_<n>.gcode entries. Re-slice and export a printable 3MF.

Check that you exported the sliced plate, not the geometry project. Inspection and slicing errors stop before upload.

Slicing recipe (FULU, OrcaSlicer, or Bambu Studio)

1. Open FULU OrcaSlicer-bambulab, OrcaSlicer, or Bambu Studio and load the mesh.
2. Pick the printer profile that matches the target machine (H2S, H2D, H2C, X1C, P1S, A1, ...). The start g-code differs per series; a plate sliced for X1 will heat-soak wrong on H2, and an H2C should not be treated as H2D.
3. Pick the filament in the slot you actually have it loaded in (AMS unit + tray). The plate's filament_ids is the lookup the MCP uses to build ams_mapping.
4. Slice the plate.
5. Export the sliced plate as something.gcode.3mf (Bambu Studio: File → Export → Export plate sliced file).
6. Hand that path to the MCP print_3mf tool.

⚠️ Avoid re-using an old Cube.gcode.3mf from a different printer/AMS setup. Stale multi-filament declarations in the file will fight the AMS mapping at print time. When in doubt, re-slice fresh.

Firmware routing (handled internally)

For the default direct LAN path, the MCP selects the upload location and command for a pre-sliced .gcode.3mf by model:

Model	Upload location	Command
P1P / P1S / X1 family / A1 mini	/cache/<file>	gcode_file
Full-size A1	SD root /<file>	project_file, file:///sdcard/<file>
P2S	/cache/<file>	project_file, ftp:///cache/<file>
H2S / H2D / H2C	SD root /<file>	project_file, ftp:///<file>
X2D	Native, to the SD card/USB stick or eMMC, on macOS or Linux	print_3mf uses a locally built helper and installed plug-in; legacy FTPS/remote-file starts are rejected

The filename suffix is significant for the legacy route: the code recognizes .gcode.3mf for gcode_file, while other .3mf names use project-file handling. Keep the slicer's sliced-file export name when using the legacy path. Do not rename unsliced content to force routing.

These routes are implemented in src/printers/bambu.ts. The FULU BambuNetwork bridge is separately selected and has its own runtime parameters; this table does not establish bridge compatibility.

AMS mapping (auto-derived from the 3MF)

The MCP reads Metadata/plate_<n>.json.filament_ids plus the ; filament_ids = … header in plate_<n>.gcode to build ams_mapping / ams_mapping2 automatically. The caller only specifies which AMS tray each project-level filament should pull from. You no longer need to hand-compute [-1, 1, -1, -1].

For a dry run, call resolve_3mf_ams_slots on the sliced 3MF. It reads Metadata/slice_info.config for each required tray_info_idx and compares those RFID-style filament ids against the live AMS inventory. If all required filaments are loaded, it returns the ams_slots array that print_3mf will accept. For printing, pass auto_match_ams: true to let print_3mf apply the same match automatically; explicit ams_slots or ams_mapping still take precedence.

Quick troubleshooting

Symptom	Likely cause	Fix
3MF does not contain any Metadata/plate_<n>.gcode	File is a mesh .3mf, not a sliced one	Re-slice and export the sliced file
405004002 on a legacy P1/X1/A1 mini route	Unexpected container/command combination	Check model, firmware, and sliced-file suffix against the routing table; full-size A1 uses a different route
0700-8012 032015	AMS-map length mismatches plate's filament count	Re-slice; don't hand-edit the file. Confirm AMS slot matches loaded filament
Print starts, heats, no extrusion	Stale start-g-code from different printer profile	Re-slice with the correct printer profile
CLI slicing fails	Missing profiles, unsupported flags, or slicer-version failure	Preserve the error; use matching presets or GUI-export a sliced .gcode.3mf. Do not upload the unsliced input