I’m currently furloughed and using a free Palantir account to do some prototyping. I have some architectural questions around integrating multi-agent systems on the Palantir platform and would appreciate a design consult with a Palantir engineer if that’s possible.
If anyone at Palantir is interested I’d be happy to share the design doc and the repo
Thanks for posting on Community!
Design Consultations are unfortunately not something we can offer for those on Developer Tier.
That being said, this question is perfect for this site. If you could describe your needs in more detail here the wider Community may be able to help.
Michael - thanks for your response.
There’s an application developed on the Palantir platform that allows multiple users to collaborate for a specific type of mission.
We want to create agents that will interact with this application in ways that emulate those human users. The agents have access to the same documentation, tools and data sources as the human users.
The fundamental design question is do we create the agents on the Palantir platform, or are there benefits to keeping them external to Palantir and have them interact with the application via an API or using MCP.
The benefit of having the agents internal to Palantir is that they would have access to more information, they could see the ontology, etc… but the downside is that it is more complex to debug and there could be less transparency in the agents actions.
On the other hand, having the agents external to the Palantir platform is structured more like the way humans interact with the application, simplifying the interface and maybe even enhancing safety and security.
I’d be interested in hearing if anyone has explored these options and what led them to go one way or the other.
Thanks,
Kevin