Dynamo 4.2 is out, and the big item is Autodesk Assistant shipping built in, with DynamoMCP alongside it.
The MCP piece is what makes it more than a chat window. Assistant can place and wire nodes, set inputs, run the graph, read warnings, group things, and save. If you ask about a node that has an authored sample graph, it inserts the actual example and adapts it instead of building something approximate from scratch.
I’ve been playing with this for a while now, and I wish I had it when I was learning Dynamo. Asking “why is this node throwing a warning” in the workspace beats endless searching.
A few practical notes:
Sandbox only for now, host integrations come later. In Revit you’ll go through Revit’s own Assistant instead of this panel.
You need to be signed in, and it’s unavailable in no-network mode.
It can’t install packages or touch the Revit document directly.
Also for the automation folks, there’s a new public interface exposing type info for Define Data nodes, so you can read a graph’s expected inputs without opening it in Dynamo. Useful for Player-style runners or validating graphs in CI.
Builds are on dynamobuilds.com or GitHub, full release notes on the wiki. Curious what people try with the Assistant first.
i see many reason for that maybe, not all is in these version, and just a really fair question “how do I export an Excel file” with ootb instead of a bot there just do it for you…but just my IMO, and maybe im totally wrong here ; and personal for me hare the way to the result always been more funny then result itself
I can’t give numbers, but I can say there’s a good bit of interest and it feels like there’s more early uptick than we see with a lot of other features (i.e. the ability to insert a sample graph for a given node via the extended node help is not well known and is in all currently supported Dynamo releases).
Actually the prompt was all me… Just robotically driven input from off screen after a few test runs with similar prompts.
This particular recording was a demo I built to show both ‘what could be’ and the scope of what the tool can do (i.e. authoring and editing with nodes, design script, python). Full prompt was written by me, but in stages as I went for trial and error finding some consistency that covered all my desired illustrations and then sent using an automation tool as if it had been typed. My prompts for other things are often similarly worded (input tokens cost less than outputs so steering the output early has measurable impact on ROI). Smaller prompts certainly work, but less consistently (i.e. 50% chance the material often gets applied as paint if you go all Python) or by skipping stuff I wanted the company I built the demo for to see (i.e. running all phases at once and skipping the watch nodes so users don’t understand how they can validate content as the graph builds).
Any particular action will have a variable time based on what else my system was up to (my recording tool and the tool used to drive the input for the demo both made things Much slower than if I didn’t record and just pasted the full prompt). In some cases adding a watch node takes a second (the smallest unit of time it reports) but in others it’s a fair bit longer. The action isn’t just find a watch node, but more akin to all of this:
Identify the need for a watch node
Confirm the use of the watch node
Find the right node
Find the approximate size of the node
Take stock of all the nodes on canvas to see where everything is
Within the context of everything on canvas find a place the node fits
Put the node at that place
Find the output port to connect to the new node
Find the correct input of the new node
Connect the identified output to the identified input
Run the graph to validate it’s output is expected and there are no new warnings or errors
Group and document the node as needed
Obviously for a simple node that’s a bit extreme, and no one wants to see 21 seconds for that, when it took 36 seconds to build the full graph leading into that watch node, but what the assistant shows is often a LOT less than what it does.
A simpler ‘first prompt’ that might get the gears turning for some of us:
Build a simplified replica of the starship enterprise.
Usually you’ll get something not quite right, but after a few corrective prompts you can get pretty good outcomes.
In about a 5 minute span and three re-prompts. My full list of subsequent prompts to get this result:
I have wired things up differently, review and then look at the relative location of the secondary hull, nacelles, and pylons in comparison to the primary hull.
yes
are all parts touchin as they should?
yes
*the neck still has a flat surface - adjust it to start at the center of the main hull, and pull the front of the secondary hull forward to align with the front ege of the nacelles
For geometry creation you can also build more complex things such as this clockwork assembly, all of which comes from AA including the animation components (the hour gear can be a bit hard to identify in the animation, but trust me it spins well). Note the watch node with a single double in it shows the geometry overlap, always within a floating point margin of error. Would need a base plate, and it might not be ready for fabrication, but I’d considering sending it to a 3D printer if I had the hardware and filament to spare.
@Andrew_Hannell those “how do I export an Excel file” threads are a good example of where this actually helps, especially if you’re just getting started. You can ask that in the workspace, in plain language, and get both an explanation and a graph you can look at. @sovitek made a fair point too, that a lot of people would rather learn the OOTB way than have a bot just do it for them. That’s still on the table here. Assistant leaves the nodes on the canvas, so you can see how it was put together, change it, and actually learn the graph.
Hey C3D users! While I cannot give any dates regarding features coming out for a host product, I can say that we are working with the appropriate teams to work toward this.
I will post a top level post here in the forum when I have more information.
and personal for me hare the way to the result always been more funny then result itself
…funny future we all go into…no matter what or like it or not 
