I hope you can help me, because I’m about to start working with AI (Codex, in my case) and I’m realising that Codex works very well for building suites, test cases and resource files without needing to configure any specific MCP for RF, Selenium or browser libraries.
Is it really important to use a specific RF MCP? Does it significantly improve interaction, or is a good AGENTS.md, Skills and Rules file sufficient?
Also, do you know of any AGENTS/Skills or Rules files that contain information or a basic outline for working with RF and the browser library?
I’m Many , the maintainer of the robot framework MCP.
I would say - if it works well enough for you in your development flow without the MCP - then I would not use it. Nowadays you can do a lot using agent skills, sub agents.
There are some things which won’t work without using the MCP Server - like a proper step wise execution which lets the agent create your test cases stepwise and will return a fully working test suite at the end. Without the MCP this will need a lot of trial and error. In some technologies (like web automation or mobile automation) this trial and error can take long - in other areas like API testing this is not a big issue.
Regarding existing agent skills - I started creating a skill set which we will develop further by the AI workgroup of the robot framework foundation. Maybe it’s already helpful enough for you - but there’ll be some more updates in the future.
All the best, looking forward to hearing your experience
Thanks, @Many.
As I’m just getting started with this (I’ve only been working with AI for six weeks), I’ll try out different configurations, with and without your MCP, and I’ll put your skills to the test as well.
Many thanks!