Bambu Studio AI

🖨️ Bambu Studio AI

The first full-pipeline AI 3D printing skill — from a chat message to a finished print.

Connect your Bambu Lab printer + pick a 3D generation API, and your agent handles everything: search → generate → analyze → repair → preview → print → monitor.

Supports all 9 Bambu Lab printers. Cloud + LAN dual mode.

Installation

clawhub install bambu-studio-ai

Or clone from GitHub:

git clone https://github.com/heyixuan2/bambu-studio-ai.git ~/.agents/skills/bambu-studio-ai
pip3 install bambulabs-api bambu-lab-cloud-api requests trimesh

📦 GitHub: https://github.com/heyixuan2/bambu-studio-ai

Quick Reference

Detection Triggers

Activate this skill when you see:

First-Time Setup

If config.json does not exist in the skill directory, walk the user through setup via conversation before running any commands.

Phase 1: Configure

Ask these in order. Be conversational, not robotic.

1. Printer Model

"Which Bambu Lab printer do you have?"

Show options grouped by series:

• 🟢 A Series (Entry): A1 Mini (180³mm, 500mm/s), A1 (256³mm, 500mm/s)
• 🔵 P Series (Prosumer): P1S (256³mm, 500mm/s, enclosed), P2S (256³mm, 600mm/s)
• 🟠 X Series (Pro): X1C (256³mm, AI features), X1E (Industrial, HEPA)
• 🔴 H Series (High-end): H2C (350°C, heated chamber), H2S (340³mm, 1000mm/s), H2D (dual extruder, laser, cutting)

2. Connection Mode

"How is your printer connected?"

🔌 LAN (Recommended) — If your computer and printer are on the same WiFi/network. Full features: camera, G-code, AMS, AI monitoring, fast response. You'll need: IP address, serial number, and access code from printer Settings → Device. Also make sure LAN Mode is turned ON on the printer (Settings → Network → LAN Mode → ON).

☁️ Cloud — If you're accessing remotely (not on the same network). Limited features: no camera, no G-code, no AI monitoring. Requires email verification.

Default to LAN unless user specifically needs remote access.

• LAN → printer IP + serial + access code (Settings → Device → LAN Access Code)
• Cloud → email + password (will require verification code on first login)
• Both → collect all, default to LAN

3. AI 3D Generation (optional)

"Want AI 3D model generation? Create printable models from text or photos."

If yes → pick provider + API key:

• Meshy — Most mature, STL/3MF, $20/mo
• Tripo3D — Python SDK, $10/mo
• Printpal — Print-optimized output
• 3D AI Studio — Early access

4. Notifications

"How should I notify you? Auto (match your channel), Discord, iMessage, Telegram, WhatsApp, or Slack?"

5. AI Print Monitoring (optional)

"Want AI monitoring? I'll photograph prints and check for failures."

If yes → interval (default 120s), auto-pause on anomaly (default no)

6. Save Configuration

Write config.json (non-sensitive, shareable):

{
  "model": "H2D",
  "mode": "cloud",
  "email": "user@example.com",
  "device_id": "",
  "printer_ip": "",
  "serial": "",
  "3d_provider": "meshy",
  "notify_channel": "auto",
  "monitor_enabled": false,
  "monitor_interval": 120,
  "auto_pause": false
}

Write .secrets.json (sensitive, use these exact keys):

{
  "password": "bambu_account_password",
  "access_code": "printer_lan_access_code",
  "3d_api_key": "generation_api_key"
}

Then run:

chmod 600 .secrets.json

Ensure .gitignore includes .secrets.json and config.json.

Phase 2: Verify (ask permission)

"Config saved! I'd like to run quick tests — no prints will start, no significant API costs. OK?"

Only proceed if user agrees.

If Bambu Studio not installed, offer: brew install --cask bambu-studio

Phase 3: Summary

🎉 Setup Complete!

🖨️ Printer:     H2D (350×320×325mm, 600mm/s)
☁️ Connection:  Cloud ✅
🎨 3D Gen:      Meshy ✅
🔍 Monitor:     ON (every 120s)
📸 Camera:      ✅
📦 AMS:         4 slots loaded
📥 Studio:      ✅ Installed

Try: "What's my printer status?" or "Generate a phone stand"

Model Sourcing (Search Before Generating)

When a user wants to print something, always ask first:

"Do you want me to:

1. 🔎 Search online — find existing models (usually higher quality, tested by community)
2. 🎨 AI Generate — create a custom model from scratch
3. 🤷 Not sure — I'll search first, generate if nothing fits"

Then follow the appropriate flow. Don't assume — the user may already have a specific design in mind, or may want something completely custom that doesn't exist.

Search Priority

When to Search vs Generate

Search Flow

1. Ask user: "Want me to search for existing models first? Often better quality than AI-generated."
2. If yes → search Thangs/Printables/MakerWorld via web_search
3. Present top 3-5 results with:
   • Name, thumbnail URL, download count/rating
   • File format (prefer .3mf > .stl)
   • Whether it has Bambu Lab print profiles
4. If user picks one → download → analyze.py → preview → print
5. If nothing good → fall back to AI generation

Example

User: "Print me a headphone stand" Agent: "Let me search — headphone stands are very common online." → Searches Printables + MakerWorld Agent: "Found 3 good options:

1. ⭐ 'Minimal Headphone Stand' — 4.8★, 12K downloads, has Bambu profile
2. 'Under-desk Hook Stand' — 4.6★, wall-mounted
3. 'RGB Headphone Holder' — with cable channel

Want one of these, or should I generate a custom design?"

Pre-Generation Flow

Before generating any 3D model, follow these steps. Do NOT skip.

Step 1: Clarify Requirements

Ask about anything unclear:

Example:

User: "Print me a phone stand"
Agent: "Sure! Quick questions:

1. Which phone? (affects holder width)
2. Viewing angle? (15°, 45°, upright?)
3. Any features? (cable hole, non-slip base?)
4. Material preference?"

Step 2: Research Unknown Objects (ask permission)

If you don't know exact dimensions or appearance:

User: "Print an iPhone 15 Pro Max case"
Agent: "I'll need the exact dimensions for a proper fit. Mind if I look them up?"

After research, confirm:

"iPhone 15 Pro Max: 159.9 × 76.7 × 8.25mm, camera bump 40 × 36mm.
I'll add 1mm tolerance per side. Sound right?"

What to research:

• Physical dimensions of real objects
• Tolerance/clearance for cases (typically 0.5–1mm)
• Standard sizes (screws, USB ports, cables)
• Reference images for complex shapes

Step 3: Confirm Before Generating

"Here's what I'll generate:

• iPhone 15 Pro Max case
• 162 × 79 × 12mm (1mm tolerance)
• Camera cutout: 42 × 38mm
• Material: TPU (flexible)
• Est. print time: ~2h

Ready to generate?"

Step 4: Generate

Before generating, tell the user:

"A quick heads-up: AI-generated 3D model quality depends mainly on two things:

1. Your 3D generation provider (Meshy, Tripo, etc.) — each has different strengths
2. How detailed your prompt is — the more specific, the better

AI-generated models are not production-ready out of the box. Always review in Bambu Studio. If the result isn't great, try a different provider or refine your prompt."

Call generate.py with a detailed, dimensions-aware prompt. The script auto-enhances prompts with printability instructions and scales to your printer's build volume.

Use --raw to skip auto-enhancement if you've crafted a precise prompt.

Supported Printers

A Series (Entry Level)

P Series (Prosumer)

X Series (Professional)

H Series (High-Performance)

Connection Modes

⚠️ Recommendation: Use LAN mode whenever possible. Cloud mode has limited features and requires email verification on every login. LAN mode is faster, has full features, and no auth headaches.

🔌 Local / LAN (Recommended)

Requirements: OpenClaw host and Bambu Lab printer on the same local network (WiFi or Ethernet).

How to enable LAN mode on your printer:

1. On printer touchscreen → Settings → Network → LAN Mode
2. Toggle LAN Mode to ON (Live)
3. Note down:
   • IP Address — shown on the network screen
   • Serial Number — Settings → Device Info
   • Access Code — Settings → Device Info → LAN Access Code

export BAMBU_MODE="local"
export BAMBU_IP="192.168.1.100"        # Your printer's IP
export BAMBU_SERIAL="01P00A000000000"   # From Device Info
export BAMBU_ACCESS_CODE="12345678"     # From Device Info

☁️ Cloud (remote access only)

Only use Cloud mode if you cannot be on the same network as the printer.

Limitations:

• ❌ No camera snapshots
• ❌ No G-code commands
• ❌ Limited AMS info
• ❌ Requires email verification code on first login (and after token expires)
• ⚠️ Slower response, depends on Bambu servers

Login includes verification code flow:

1. Script sends login request
2. Bambu emails you a verification code
3. Enter the code when prompted (script will wait patiently — no auto-retry)
4. Token is cached for 24 hours

export BAMBU_MODE="cloud"
export BAMBU_EMAIL="you@email.com"
export BAMBU_PASSWORD="password"

Scripts auto-load config.json + .secrets.json from the skill directory. Environment variables override config file values.

Configuration Files

config.json (non-sensitive, shareable)

See config/config.example.json for template.

.secrets.json (sensitive, chmod 600)

See config/.secrets.example.json for template. Exact keys: password, access_code, 3d_api_key

.gitignore

Ships with the skill. Excludes .secrets.json and config.json.

3D Generation

Supports 4 providers:

Auto Prompt Enhancement

User says "a phone stand" → script auto-adds:

"...Optimized for FDM 3D printing. Max 230×230×230mm. Flat base, watertight mesh, no overhangs >45°, min 1.5mm walls."

Auto Size Limiting

Models auto-constrained to printer build volume (10% margin):

Skip enhancement: --raw flag.

Output Format Priority

Always generate in Bambu Lab compatible formats, in this order:

Default --format 3mf unless user specifies otherwise.

Mandatory Pre-Print Pipeline

NEVER send a model directly to the printer. Always follow this order:

1. Generate/Download → model.3mf
2. Analyze + Repair  → python3 scripts/analyze.py model.3mf --repair --material PLA
3. Report to user    → "Score 8/10, repaired 58K non-manifold edges..."
4. Open in Bambu Studio → open -a "BambuStudio" model.3mf
5. ⚠️ MANDATORY: Tell user to inspect in Bambu Studio
   → "I've opened the model in Bambu Studio. Please check:
      - Does it look correct? Any missing or deformed parts?
      - Any floating/disconnected pieces? (they will fall during printing!)
      - Is the size right? (check dimensions in bottom bar)
      - Any red warnings? (non-manifold, intersecting parts)
      - Are there parts hanging in the air that need supports?
      - Slice it and check: estimated time, filament usage, and support amount.
      Let me know when you're ready to print!"
6. WAIT for explicit user confirmation → "looks good" / "print it" / "go ahead"
7. Print             → python3 scripts/bambu.py print model.3mf

⛔ NEVER skip step 5-6. NEVER auto-print without user seeing the model in Bambu Studio. AI-generated models frequently have mesh errors (non-manifold edges, holes, intersections). The user MUST visually verify before printing.

The 11-Point Printability Check

Also checks: watertight mesh, manifold geometry, build volume fit.

Output includes recommended print settings (layer height, infill, walls, temps, supports).

See references/3d-prompt-guide.md for detailed prompt engineering tips.

Reference Documents

The references/ folder contains docs the agent can consult:

• bambu-mqtt-protocol.md — MQTT topics, commands, report fields
• bambu-cloud-api.md — Cloud API Python SDK methods
• 3d-generation-apis.md — Meshy/Tripo/Printpal/3DAI endpoints
• 3d-prompt-guide.md — Prompt engineering for printable 3D models
• model-specs.md — All 9 printer specs in one table

AI Print Monitoring

⚠️ Requires User Consent

Always ask before enabling:

"Want AI print monitoring? I'll photograph your print and use AI to check for failures."

If yes, ask monitoring intensity:

"How closely should I monitor? This affects token usage:

• 🟢 Light (every 30 min) — ~2 tokens/hr, good for long prints
• 🟡 Standard (every 5 min) — ~12 tokens/hr, recommended
• 🔴 Intensive (every 2 min) — ~30 tokens/hr, for critical prints or new materials
• ⚫ Off — no monitoring

Which level?"

Then: "Should I auto-pause if I detect something serious like spaghetti or bed detachment?"

Never auto-enable.

Anomaly Detection

Flow

monitor.py captures snapshot → outputs image path
    ↓
Agent analyzes image (via image tool)
    ↓
Normal → log, continue
Suspicious → shorten interval, watch
Critical → notify user + optional auto-pause

Material Guide

Standard (All models, open or enclosed)

Engineering (Enclosed models: P1S, P2S, X1C, X1E, H2C, H2S, H2D)

High-Temp (H2C, H2D only — 350°C nozzle)

H2D-Specific Features

Dual Extruder

• Left (T0): Primary | Right (T1): Secondary
• Uses: dual-color, PLA+PVA soluble support, soft+hard combos
• Combine with AMS for more colors

Laser Modules

• 10W: Engrave wood, leather, acrylic
• 40W: Cut wood, acrylic sheets

Cutting Module

• Precision cutting of thin materials

Troubleshooting

Connection

Print Failures

Automation Ideas

• Progress updates — Cron push to user's channel every 30 min
• AMS low filament alert — Warn below 20%
• Timelapse — Snapshot every 2 min → ffmpeg video
• Print complete notification — Alert + final photo
• Anomaly detection — AI analysis → auto-pause + alert

Bambu Studio

Check: brew list --cask bambu-studio 2>/dev/null

Not installed? Offer:

"Bambu Studio gives live camera and slicing. Want me to install it?"

brew install --cask bambu-studio

Other platforms:

• Windows: https://bambulab.com/en/download/studio
• Linux: https://github.com/bambulab/BambuStudio/releases

Speed Modes

python3 scripts/bambu.py speed silent     # Quiet (night)
python3 scripts/bambu.py speed standard   # Standard
python3 scripts/bambu.py speed sport      # Fast
python3 scripts/bambu.py speed ludicrous  # Max (H2S: 1000mm/s)

Known Limitations

Version History

0.20.0

• NEW: slice.py — CLI slicer with OrcaSlicer backend
  • Auto-resolves Bambu Studio profile inheritance chains
  • Auto-detects printer model, nozzle size from live printer
  • Quality presets: draft/standard/fine/extra
  • Auto-orient + auto-arrange flags
  • 3MF post-processing for BS compatibility
  • All 9 Bambu Lab printer models supported
• NEW: bambu.py info — printer hardware info (model, nozzle, AMS filaments)
• NEW: bambu.py notify — send notifications
• NEW: Smart print monitor with anomaly detection
  • Progress stall detection (>10min)
  • Temperature anomaly alerts
  • 30-min progress summaries, immediate anomaly alerts
  • Auto-pause on critical anomalies
• Fixed: bambu.py dispatch for info/notify/status --json
• Fixed: generate.py --format 3mf now correctly produces .3mf
• Fixed: colorize.py auto-adds .obj extension
• Fixed: doctor.py checks OrcaSlicer installation
• Fixed: Relative output paths in slice.py
• Fixed: 3MF compatibility with Bambu Studio 2.5.0

0.19.2

• Fixed: _safe_split() with 30s SIGALRM timeout for mesh.split() hang protection
• Verified with stress test model T12 (Eiffel Tower lattice)

0.19.1

• 17 bug fixes from 3-round audit (B1-B17, R1-R4, D1-D5, O3-O6)

• search.py rewritten to use ddgs package (DDG HTML scraping broken)

• analyze.py stdlib imports moved to top

• Monitor --status works offline

• 0.19.0 — 13 bug fixes, doctor.py, --confirmed safety gate, color pipeline overhaul (AO delight, shadow-aware mapping, border vote), search rewrite (ddgs), monitor retry logic

• 0.18.0 — Model search (4 sources), notification system

• 0.17.x — Bambu Lab official 43-color palette, direct CIELAB NN (K-means removed), default LAN mode

• 0.16.x — Unit detection fix, start_print fix, MTL color fixes, OBJ meter→mm fix

• 0.15.x — OBJ mtllib reference fix after rename

• 0.14.x — Multi-color AMS: mesh-aware delight, CIELAB mapping, texture smoothing, island cleanup

• 0.13.x — GLB→STL/OBJ+MTL conversion, auto-orient, auto-scale

• 0.12.x — AI 3D generation (4 providers), prompt enhancement

• 0.11.0 — Multi-color AMS printing (colorize.py), auto-detect AMS colors

• 0.10.0 — 11-point analysis, auto mesh repair, requirements.txt

• 0.9.0 — Cloud login: token caching (24h), verification code patience, LAN recommended

• 0.8.x — Model sourcing (search before generate), user choice

• 0.7.x — analyze.py, model requirements table, security scan fixes, README

• 0.6.0 — Monitor intensity levels, 3MF priority, mandatory Bambu Studio preview

• 0.5.x — Pre-generation research flow, .secrets.example.json

• 0.4.0 — 3-phase setup (configure → test → summary)

• 0.3.x — Security fixes, config/secrets separation

• 0.2.0 — Full 9-model support, Cloud+Local dual mode, AI monitoring

• 0.1.0 — Initial release

License

MIT