Skip to content

PyNET FAQs

Jump to bottom
Rafael Núñez de Arenas edited this page Apr 18, 2026 · 6 revisions

Frequently Asked Questions

What is PyNET Platform?

PyNET Platform is an AI-driven BIM automation ecosystem that connects natural language, Python execution, and Autodesk desktop tools into a unified workflow.

It allows AI models to generate, execute, and refine Python scripts directly inside Navisworks, Revit, and AutoCAD/Civil 3D.

Result: automate repetitive BIM tasks faster and reduce manual scripting effort.

Who is PyNET Platform for?

PyNET is designed for:

  • BIM managers
  • Technical architects and engineers
  • Automation-focused teams
  • Developers working with Autodesk APIs

If you want to reduce repetitive work in Navisworks or Revit, PyNET is built for you.

Are there any tutorials available to learn how to use the tool?

Yes. We have a dedicated video series to help you get up to speed. You can find the full list of tutorials here:

Navisworks PyNet Guide - YouTube Playlist

Which Autodesk products are supported?

Currently supported:

  • Autodesk Navisworks Manage

Coming soon:

  • Autodesk Revit
  • Autodesk AutoCAD / Civil 3D

You can start automating today in Navisworks and expand as new integrations are released.

What is PyNet Library?

PyNet Library is a reference repository containing Python-style API stubs and example scripts for Autodesk Navisworks.

It provides AI models (and developers) with the context they need to generate accurate automation scripts, including type hints, method signatures, boilerplate templates, and working examples organized by use case.

What is PyNet Bridge?

PyNet Bridge is the MCP (Model Context Protocol) server that connects AI models to PyNET Platform.

It acts as the communication layer between your AI client and the Autodesk application, handling:

  • Script validation
  • Execution requests
  • UI management
  • Real-time feedback

How do I install PyNet Bridge?

Before you start, make sure you have:

  1. Python 3.10+ installed (up to 3.13 — see What version of Python do I need?) — python.org
  2. Git installed — required for VS Code extensions (Claude Code, Cline, Roo Code) → git-scm.com
  3. At least one AI client installed — the installer only configures clients it finds on your system, so install your preferred client first (Claude Desktop, VS Code with Claude Code/Cline/Roo Code)

Then open PowerShell and run:

irm https://raw.githubusercontent.com/rafa2403nunez-droid/PyNetBridge/main/install.ps1 | iex

The installer will automatically install the MCP server and configure all detected AI clients.

Tip: Restart your AI client(s) after installation.

For manual installation, see the PyNet Bridge README.

Do I need an AI subscription?

Not necessarily.

PyNET Platform does not include access to any AI provider, but you have two options:

  • Cloud AI (paid) — Use providers like Claude, OpenAI, Gemini, or DeepSeek with an API key
  • Local AI (free) — Run models locally via Ollama or LM Studio

You can also use PyNET Platform without AI as a standalone Python scripting engine.

Which AI clients are compatible?

The PyNet Bridge (MCP) server supports:

  • Claude Desktop (standard and Microsoft Store)
  • Claude Code (VS Code extension / CLI)
  • Cline (VS Code extension)
  • Roo Code (VS Code extension)

Any MCP-compatible client that supports the stdio transport should work.

How do I install a compatible AI client?

Client Type Download
Claude Desktop Standalone app claude.ai/download
Claude Code VS Code extension VS Code Marketplace
Cline VS Code extension VS Code Marketplace
Roo Code VS Code extension VS Code Marketplace

Note: Claude Code, Cline, and Roo Code require Visual Studio Code and Git to be installed first.

Which AI API Keys can I get for free?

If you are looking to build applications, experiment with AI coding agents like Cline or Roo Code, or use MCP (Model Context Protocol) without an upfront cost, these are the best providers in 2026.

1. Google AI Studio (Gemini)

Google currently offers the most generous free tier for developers, especially regarding context window size.

  • Best for: Massive context (up to 1M+ tokens). Ideal for Cline/Roo Code to read your entire project at once.
  • Models: Gemini 2.5 Flash, Gemini 2.0 Flash, Gemini Pro.
  • Limits: High Rate Limits (RPM) for Flash models.
  • Get it here: Google AI Studio

2. Groq Cloud

Groq is famous for its extreme inference speed using specialized LPU hardware.

  • Best for: Instant responses. Makes Roo Code feel incredibly snappy when generating small snippets.
  • Models: Llama 4, Llama 3.1 (70B/405B), Mistral, DeepSeek.
  • Limits: Rate-limited based on tokens per minute (TPM).
  • Get it here: Groq Console

3. Mistral AI

The premier European AI lab offers an "Exploration" tier for their high-reasoning models.

  • Best for: High-quality reasoning and multilingual support.
  • Models: Mistral Large, Mistral NeMo, Codestral (specifically built for code).
  • Limits: Usage-based limits for non-commercial/testing purposes.
  • Get it here: Mistral La Plateforme

4. OpenRouter (Free Models)

OpenRouter acts as a unified interface. They maintain a specific list of models that are 100% free provided by various hosts.

  • Best for: Flexibility. You can switch between models in Cline/Roo Code without changing API keys.
  • Models: DeepSeek V3 (Free), Qwen 2.5 Coder, Phi-3.
  • Note: Look for models tagged as "FREE" in their ranking list.
  • Get it here: OpenRouter

5. Hugging Face (Inference API)

Access thousands of community-uploaded models through a simple API.

  • Best for: Specific use cases like Image-to-Text, Audio, or specialized NLP.
  • Models: Virtually any Open Source model (Llama, Falcon, Stable Diffusion).
  • Limits: Shared infrastructure; speeds may vary depending on community load.
  • Get it here: Hugging Face Settings

Summary Table for Developers

Provider Top Benefit Best Model for Coding (Cline/Roo)
Google 1M Context Window Gemini 2.5 Flash
Groq Speed (Low Latency) Llama 4
Mistral Reasoning Quality Codestral
OpenRouter Variety / No Lock-in DeepSeek V3

Pro-Tip: For agents like Cline and Roo Code, we highly recommend starting with Gemini 2.5 Flash. Its ability to process huge codebases (context window) is a game-changer when using MCP servers to analyze multiple files or logs.

Can I use PyNET with a free or local AI model?

Yes. You don't need a paid AI subscription to use PyNET Platform with AI.

Cline and Roo Code support local AI models through providers like:

Provider Description Download
Ollama Run open-source models locally ollama.com
LM Studio Desktop app for local models lmstudio.ai

Any provider with an OpenAI-compatible API works. Popular free models include Qwen, Llama, DeepSeek, and Mistral.

Setup: Install Ollama or LM Studio, download a model, then configure Cline or Roo Code to point to your local endpoint (e.g. http://localhost:11434 for Ollama).

The PyNet Bridge MCP server works the same regardless of which AI model is behind the client.

Important: Running local AI models requires significant hardware. Minimum recommended specs:

  • RAM: 16 GB (32 GB recommended)
  • GPU: 8 GB VRAM (e.g. NVIDIA RTX 3060/4060 or higher)
  • For 30B+ parameter models: 16+ GB VRAM (e.g. RTX 3090/4080/4090)

Without a dedicated GPU, models will run on CPU only, which is significantly slower. Larger models generally produce better Navisworks scripts.

How do I configure Cline or Roo Code?

Both Cline and Roo Code work the same way — they are VS Code extensions that connect to an AI provider of your choice.

On first launch, they ask you to configure a provider.

Option A: Cloud AI providers (API key required)

Provider Get API Key Notes
Anthropic (Claude) console.anthropic.com Pay-per-use, high quality
OpenAI (GPT) platform.openai.com Pay-per-use
Google (Gemini) aistudio.google.com Free tier available
DeepSeek platform.deepseek.com Low cost

In Cline/Roo Code settings, select your provider and paste your API key.

Option B: Local AI (free, no API key)

  1. Install Ollama or LM Studio
  2. Download a model (e.g. ollama pull qwen2.5-coder:32b)
  3. In Cline/Roo Code settings, select "Ollama" or "OpenAI Compatible" as provider
  4. Set the endpoint to http://localhost:11434 (Ollama) or http://localhost:1234 (LM Studio)

No API key needed — everything runs on your machine.

After configuration

Once configured, PyNet Bridge MCP tools will appear automatically in the client (if installed). The AI can then interact with Navisworks directly.

What version of Python do I need?

Python 3.10 or higher is required.

Install it from:

⚠️ Python 3.14 is not yet supported. The pythonnet runtime currently supports Python 3.7 through 3.13. If you encounter a System.NotSupportedException mentioning an unsupported ABI version, install Python 3.12 or 3.13 and configure PyNet to use it.

Which Python libraries should I install?

PyNET Platform works out of the box with standard Python, but you can install a recommended set of libraries to enable more advanced scripting capabilities.

Python Libraries Starter Pack (optional)

Install the recommended libraries with:

irm https://raw.githubusercontent.com/rafa2403nunez-droid/PyNetBridge/main/install-libraries.ps1 | iex

This installs:

Library Purpose
pandas Data analysis and manipulation
plotly Interactive charts and visualizations
matplotlib Static plots and graphs
dash Build web dashboards with Python

These libraries are included in the list of allowed Python imports for secure execution.

Standard library modules (such as json, sys, re, etc.) are already included with Python and do not need to be installed.

Prerequisites

Is my BIM data sent to the cloud?

No.

All script execution happens locally on your machine. PyNET Platform does not transmit any project data to external servers.

However, when using AI integration, prompts and generated scripts are subject to your AI provider’s privacy policy.

Can I create my own scripts?

Absolutely.

PyNET Platform includes a full embedded Python.NET engine. You can:

  • Write custom scripts
  • Interact with the Autodesk API
  • Create UI buttons linked to your scripts
  • Build complete automation workflows

See the PyNet Library for reference scripts.

Where are the processes executed by PyNet saved?

PyNet develops and executes code directly within the target software (such as Navisworks) for immediate results. By default, these temporary scripts are not saved as permanent files. However, if you want to keep the code, you can simply ask the AI to save the script into your specific folder structure, and it will generate the file for you.

Keep in mind that the user has total control over where these files are stored. Regarding file writing, AI clients will always ask for your explicit permission before generating or modifying any files; they will never create them without your approval.

How do I give AI context about the Autodesk API?

The PyNet Library provides all the context AI models need to generate accurate scripts.

It includes API stubs, example scripts, and boilerplate templates.

Add the repository as context in your AI client. The included CLAUDE.md file provides structured instructions automatically read by the model.

How does the security validation work?

The PyNet Bridge (MCP) includes a built-in static analyzer that validates every AI-generated script before execution.

Scripts that violate the rules are rejected immediately.

Validation includes:

  • Allowed CLR assemblies
  • Allowed Python imports
  • Blocked imports (e.g., os, subprocess, socket)
  • Blocked calls (eval, exec, __import__)
  • Blocked attributes (__builtins__, __globals__, etc.)

See the full rules in the PyNet Bridge README.

How do I get a PyNET Platform license?

PyNET Platform is distributed through Freemius.

Purchase and manage your license here:
Freemius Store

Where do I report issues or get support?

Report bugs or request features on GitHub Issues:

Clone this wiki locally