pyjal3dsmcp
Connects to a running Autodesk 3ds Max instance on a Windows workstation, letting an assistant inspect the scene (e.g. scene overview and object counts), find objects by name, and work with materials and modifiers. It also provides tools for discovering, registering, and executing custom Python libraries inside Max, with per-instance approval of imports and execution.
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
PyJal3DSMCP
Connect Claude Desktop, Codex, or ChatGPT Desktop to Autodesk 3ds Max on your Windows workstation. Ask your assistant to inspect a scene, find objects by name, or work with materials and modifiers. PyJal3DSMCP includes the tools from 3dsmax-mcp and adds 12 tools for discovering and running your own Python libraries inside Max.
Ask an agent to install it
Copy this into an agent that can run PowerShell on the PC where you use Max. Replace Claude Desktop with Codex or ChatGPT Desktop if that is the app you use.
Install https://github.com/jalnaga/PyJal3DSMCP for Claude Desktop on this Windows PC. Follow the README and register the server for that app. Let me save my scene if Max needs to close. After restart, verify the connection by reading a scene overview.
You need Windows, 3ds Max 2023–2027, and a desktop client with local MCP support. For ChatGPT Desktop, use a local Work session with Settings → MCP servers. The client launches the server locally over standard input/output (stdio). OpenAI desktop MCP setup
Related MCP server: MaxForge MCP
Installation and updates
On Windows, install uv and Git and choose the full 40-character commit of the public release. The isolated launcher prepares its own cache. Review the plan, save open scenes and close any apps listed as holding a lock before applying it.
$sha = ((git ls-remote https://github.com/jalnaga/PyJal3DSMCP HEAD) -split '\s+')[0]
if ($sha -notmatch '^[a-f0-9]{40}$') { throw 'Cannot resolve release commit' }
$source = "git+https://github.com/jalnaga/PyJal3DSMCP@$sha"
uvx --isolated --from $source pyjal3dsmcp-install --status --json
uvx --isolated --from $source pyjal3dsmcp-install --client claude-desktop --dry-run --json
uvx --isolated --from $source pyjal3dsmcp-install --client claude-desktop --yes --jsonRepeat --client to select codex or chatgpt-desktop. Those two clients share one TOML entry. Use --all-clients to discover all supported clients, and --all-clients --no-apps for CLI registration only. Unattended runs skip skills by default. Select --skill global or --skill project explicitly. If the plan requires administrator access, review the destination and use --allow-elevation to permit the Windows approval dialog.
$plan = uvx --isolated --from $source pyjal3dsmcp-update --ref master --client claude-desktop --dry-run --json | ConvertFrom-Json
$plan
$reviewed = $plan.build.target.commit
uvx --isolated --from $source pyjal3dsmcp-update --ref $reviewed --track-ref master --client claude-desktop --yes --json
uvx --isolated --from $source pyjal3dsmcp-uninstall --dry-run --json
uvx --isolated --from $source pyjal3dsmcp-uninstall --yes --jsonExit 0 means complete, 1 means an operation failed, 2 means preflight blocked the operation, and 3 means follow-up is required. An uninstaller running inside the installed tool may return an after_exit command. Execute it after the process exits, then check status from the isolated launcher. Normal removal preserves personal settings. Add --purge only to remove local settings and the catalog.
Git-free release bundle
For Windows x64 with the Python version listed in the release, compare the ZIP SHA-256 with the digest in the release notes, extract it into a persistent local directory and run these commands there. For an update use the newly verified bundle and repeat the plan and apply. For removal use this bundle's isolated launcher.
$wheel = (Get-Item .\pyjal3dsmcp-*.whl).FullName
$env:UV_PYTHON = '3.13'
$env:UV_OFFLINE = 'true'
$env:UV_NO_INDEX = 'true'
$env:UV_FIND_LINKS = [uri]::new((Get-Location).Path).AbsoluteUri
$env:UV_OVERRIDE = (Get-Item .\overrides.txt).FullName
uvx --isolated --from $wheel pyjal3dsmcp-install --status --json
uvx --isolated --from $wheel pyjal3dsmcp-install --wheel $wheel --client codex --dry-run --json
uvx --isolated --from $wheel pyjal3dsmcp-install --wheel $wheel --client codex --yes --json
uvx --isolated --from $wheel pyjal3dsmcp-update --wheel $wheel --client codex --dry-run --json
uvx --isolated --from $wheel pyjal3dsmcp-update --wheel $wheel --client codex --yes --json
uvx --isolated --from $wheel pyjal3dsmcp-uninstall --dry-run --json
uvx --isolated --from $wheel pyjal3dsmcp-uninstall --yes --json
uvx --isolated --from $wheel pyjal3dsmcp-install --status --jsonPersonal and team profiles
Create an empty profile in your own directory or connect a team profile that your team already maintains. The environment variable PYJAL_DLEL_TEAM_PROFILE overrides this PC's saved pointer. Inspect the effective path and signature with show.
$file = "$HOME\MaxLibraries\personal.toml"
New-Item -ItemType Directory -Force "$HOME\MaxLibraries"
pyjal3dsmcp-team init $file --anchor "$HOME\MaxLibraries"
pyjal3dsmcp-team validate $file
pyjal3dsmcp-team set $file
pyjal3dsmcp-team show --json
pyjal3dsmcp-team edit --file $file --personal-file trust-root add sample --path "$HOME\MaxLibraries\sample"Trust-root changes grant permission to import and execute libraries beneath the root. Review the diff, then append --apply --expect-sha256 <plan hash> to the same edit command. For a team profile replace --personal-file with --work-copy <path> and use a copy with the same relative directory structure. The team maintainer reviews and publishes that copy. After a change, restart the MCP server, run query_scene(action="overview") and compare library_attach(expected_pid=PID) with the signature from show. To disconnect only the saved PC pointer, run pyjal3dsmcp-team clear.
The first ordinary library_* call and each library_attach promote old registrations whose ID and resolved source path match a discovered trusted surface. The next successful attach reports their IDs once in auto_migrated. Before migration, use library_diagnose_surfaces to compare registered and discovered paths without writing the catalog or connecting to Max.
Verify from your app
Start Max and open a new conversation after reloading the client. The server name is pyjal3dsmcp.
Client | Where to check |
Claude Desktop | Fully quit and reopen Claude. Open Settings → Developer → Edit Config to see the registered command. The installer handles both classic and Microsoft Store config locations. Claude Desktop MCP setup |
Codex | Open a new local session and type |
ChatGPT Desktop | Open Settings → MCP servers, enable the server if needed, select Restart, then type |
Use PyJal3DSMCP to list running 3ds Max instances. If there is one, read its scene overview and tell me the object count. If there are several, ask which one I want. Leave the scene as it is.
A successful query_scene call with action="overview" confirms the connection all the way to Max. If the server is listed but no Max instance appears, check the installer's output for your year's mcp_bridge_YYYY.gup and restart Max. If the server is missing, check that the app is using the config path printed by the installer. For Codex and ChatGPT this defaults to %USERPROFILE%\.codex\config.toml and follows CODEX_HOME when set.
Using Python libraries inside Max
For ordinary scene work, use the dedicated Max tools. For a custom library, select the intended Max PID, run library_attach, then use library_search and library_describe to inspect the API before library_execute. Register personal libraries with library_register and approve import and execution per instance. The server uses the full tool profile. The installer sets safe_mode = false in a new Max configuration to permit Python execution and respects an existing setting.
Teams can put pyjal3dsmcp-team.toml in the installer's working directory to register shared libraries and trusted roots. PYJAL_DLEL_TEAM_PROFILE overrides that profile location. Libraries under a trusted root inherit import and execution approval. The library agent guide (Korean) is also available as the MCP resource resource://pyjal3dsmcp/agent-guide.
License
MIT License. The upstream dependency is distributed under its own license.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
Control Unreal Engine to browse assets, import content, and manage levels and sequences. Automate…
AI Hub for AEC — 50+ 3D formats, clash detection, ACC integration via Autodesk Platform Services.
Connect your AI assistant to Blender for 3D modeling, Geometry Nodes, materials, lighting, animation, rendering and export. Inspect real scenes and viewport images, refine the result and restore checkpoints. OAuth connects your chat to Blender Bridge Desktop on an online Windows 10/11 x64 computer running Blender 4.2+. The chat can run on another device. Requires an active Blender Bridge account; 7-day trial available. AI model access is separate.
Blender-as-a-service for agents: search 3D assets, run Blender Python, or brief the studio agent.
Related MCP Servers
- AlicenseAqualityDmaintenanceEnables AI assistants to directly control Autodesk 3ds Max through natural language to create models, set materials, adjust lighting, and automate animations. It bridges the Model Context Protocol with 3ds Max via a TCP socket to execute Python and MAXScript commands.2418MIT
- FlicenseBqualityCmaintenanceEnables AI agents to control a live 3ds Max session directly, including creating and editing objects, materials, animation, and scene inspection, with undo support.17-
- AlicenseNot gradedqualityBmaintenanceEnables AI agents to connect to Autodesk 3ds Max via the Model Context Protocol, allowing natural-language control of scene creation, modeling, materials, viewport capture, and plugin introspection through dedicated MCP tools.MIT
- AlicenseAqualityCmaintenanceEnables MCP clients and AI assistants to inspect and modify a live Autodesk 3ds Max session — reading the selection, querying objects and layers, capturing the viewport, and running MAXScript or Python in the open scene. Every request is logged and each scene change becomes a named undo record, so the AI can only undo its own steps and the user can audit and reverse everything.10MIT