Skip to main content
Glama

anycubic-mcp

Unofficial Anycubic cloud 3D-printer MCP server. Send print jobs, check the bed with the printer camera + AI vision, manage a queue, query print history — all through Model Context Protocol tools, from any MCP client (Claude Desktop, Cursor, dsh, ...).

Works with printers that use the Anycubic cloud platform (verified on Kobra X with ACE multi-color; same API family should apply to other cloud models — reports welcome).

⚠️ Unofficial, reverse-engineered, MIT-licensed. Not affiliated with Anycubic. The platform may change its protocol at any time. Use at your own risk.

Features

  • ✅ Cloud print pipeline — upload G-code, start/cancel prints, live status, temps, firmware

  • ✅ Camera bed-check — Agora RTC puller (native, no resident process) + strict true/false AI vision verdict; refuses to print on a dirty bed

  • ✅ Safety prechecks — busy check → vision (authoritative) → server clean-plate flag (fallback); duplicate guard via history

  • ✅ Queue — cloud farm bulk-task API where available, local JSON queue as fallback, timer-friendly queue_process

  • ✅ Print history — merged view so agents can tell printed from unprinted

  • ✅ One-command setup — anycubic-setup scans a logged-in Anycubic app for your token, lists printers, writes config

Related MCP server: bambu-cloud-mcp

Quick start

pipx install anycubic-mcp          # or: uv tool install anycubic-mcp
anycubic-setup                     # interactive: token -> printer -> config

Then register in your MCP client:

{
  "mcpServers": {
    "anycubic": { "command": "anycubic-mcp" }
  }
}

Camera bed-check additionally needs kx_cam.exe, built once from native/ (Agora C++ SDK download instructions included), and any OpenAI-compatible multimodal endpoint for the vision verdict (local Ollama works).

Tools

Tool

Purpose

printer_status

busy/idle, temps, clean-plate flag, firmware

bed_photo

capture one frame; optional AI verdict true/false

verify_gcode

first-layer scatter + tool-sequence check — catches stacked plates & wrong colors before sending

upload_gcode

upload a file, returns cloud_file_id

send_print_job

prechecks → upload → start (the safe default); dry_run=True returns the checklist without sending

cancel_print

cancel current job

print_history

recent jobs to avoid duplicates

queue_add/list/remove/process

local queue (timer-friendly)

Slot numbering (1-based, human semantics)

slots everywhere means human numbering: [1] = first spool slot, [3] = third. The 0-based wire format is an internal detail — every send/queue result echoes a slot_map so you can see exactly which spool was used. This exists because a 0/1-based mix-up caused two real misprints.

dry_run — look before you send

send_print_job(..., dry_run=True) runs every gate (busy / vision / duplicate history / first-layer verify) and returns a checklist. Nothing is uploaded or started. Point it at any new G-code before committing filament.

Documentation

Roadmap

  • PrinterAdapter protocol: the queue executor talks to a printer only via status/upload/start/cancel/history. Anycubic cloud is the first adapter; Bambu/Klipper/Moonraker adapters can reuse the same queue, safety checks (vision bed-photo veto) and MCP surface as-is.

  • Frida-assisted capture of the farm createBulkTask body (community help welcome) — the local queue already covers single-printer workflows.

License

MIT. The signature salts are extractable from any official client binary; nothing here contains credentials. Keep your tokens in ~/.anycubic-mcp/.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      B
      quality
      B
      maintenance
      Enables interaction with 3D printer management systems through an MCP server, supporting STL file manipulation, slicing, and control of printers like OctoPrint, Klipper, Duet, and more.
      30
      5,162 npm
      266
      GPL 2.0
    • A
      license
      Not graded
      quality
      B
      maintenance
      Enables AI agents to manage multiple 3D printers via MCP, including monitoring temperatures and progress, uploading and slicing models, controlling prints, and viewing cameras.
      MIT