@lucianmincu@Alexweg3@c18365857@MultiversX They weren’t called skills 6 months ago, but that’s exactly how they worked, dynamic context building based on an open source GitHub expandable knowledge base library for MvX.
xPilot v2.0.0 is out
- new instructions library for GitHub - creating and expanding the ecosystem knowledge base for AI coding agents. Anyone can contribute in simple markdown files.
- new workflows library for GitHub - creating one shot AI executable actions for complex or repetitive project tasks.
- streamlined user interface, new AI providers, optimizations and more...
Works with VS Code, Cursor, Windsurf, or any VS Code base IDE.
Lots to add to the knowledge base, ideally it can be a community collaboration.
Any builder can add instructions about their own projects, libraries, tools, so anyone can build on them using AI.
Instruction and workflows repos could be hosted on an official @MultiversX or @xAllianceDAO organization.
xPilot v2.0.0 is out
- new instructions library for GitHub - creating and expanding the ecosystem knowledge base for AI coding agents. Anyone can contribute in simple markdown files.
- new workflows library for GitHub - creating one shot AI executable actions for complex or repetitive project tasks.
- streamlined user interface, new AI providers, optimizations and more...
Works with VS Code, Cursor, Windsurf, or any VS Code base IDE.
@HeliosStaking@MultiversX xPilot’s purpose is to bring together the ecosystem knowledge, from builders and contributors, with the intelligence of AI.
To accelerate development and allow new builders to hit the ground running.
The hackathon was the proof of concept.
Now, the link is established. 🫡
3/ xPilot
An AI software agent for blockchain developers.
xPilot connects with GPT-4.1, Claude, and Gemini to plan, write, and test code inside VS Code or other environments.
Useful for repetitive tasks or prototyping.
https://t.co/FuDj0isAud
@SasuRobert@MultiversX The mx-api mcp server is open source. If anyone needs a custom api integration for mcp compatible agents (so they can get and send data when necessary), drop us a message.
https://t.co/QcV5nsgctL
New tools for real-time natural language interaction with the @MultiversX blockchain coming soon to @xPilotAI (and any MCP compatible AI agent - Cursor, Windsurf, etc.). 🛠️
@beniaminmincu Redesigning the knowledge base implementation for the agent, so everyone can contribute and integrate their tools and knowledge for everyone else to use.
https://t.co/v5XtXmTiPi
This is actually the vision with @xPilotAI, to have a dedicated software engineering agent for MultiversX, that will allow new builders to have a starting point and to hit the ground running with appropriate guidance, not just documentation to read.
For now, this project is just one person, @x413x5, and I'm doing this in the extra time I have, with the purpose of reaching a point where the agent knowledge base can be expanded very easily by anyone in the ecosystem.
While I started using MCP for the knowledge base, I realized that is not the best way to go forward, if the goal is for everyone to contribute as easily as possible.
Recently I changed that approach, using the concept of instruction files, so that the knowledge base can be written in simple text files (markup). MCP will remain only for tools connecting to outside data (APIs, etc)
I am close to releasing a new version that will use this simplified instructions file system, with the knowledge base being stored inside a Git repository, where anyone can contribute and benefit from all the contributions, basically teaching the AI model what they know, and how to accomplish things. 🙂
There's more to it than this, if you are interested in contributing, or stay up to date with things, there is this discord server https://t.co/vghTnodpRV
This is actually the vision with @xPilotAI, to have a dedicated software engineering agent for MultiversX, that will allow new builders to have a starting point and to hit the ground running with appropriate guidance, not just documentation to read.
For now, this project is just one person, @x413x5, and I'm doing this in the extra time I have, with the purpose of reaching a point where the agent knowledge base can be expanded very easily by anyone in the ecosystem.
While I started using MCP for the knowledge base, I realized that is not the best way to go forward, if the goal is for everyone to contribute as easily as possible.
Recently I changed that approach, using the concept of instruction files, so that the knowledge base can be written in simple text files (markup). MCP will remain only for tools connecting to outside data (APIs, etc)
I am close to releasing a new version that will use this simplified instructions file system, with the knowledge base being stored inside a Git repository, where anyone can contribute and benefit from all the contributions, basically teaching the AI model what they know, and how to accomplish things. 🙂
There's more to it than this, if you are interested in contributing, or stay up to date with things, there is this discord server https://t.co/vghTnodpRV