Speelka Announcement

While I have some free time, I built a universal agent that functions as an MCP server, executes a specific prompt, and allows integration of tools from multiple MCP servers. The idea is to use it to develop standalone agents as well as entire hierarchies of agents. It is still in deep alpha with only 12% test coverage, but the core functionality is operational. Based on it, I’ve created an architect for Cursor with a narrowly focused prompt for writing ADRs, and I like it so far (though it still lacks advanced search tools like LSP).

Next, I plan to create a minimally portable agent artifact from it, enabling interaction among groups of agents in various formats. Essentially, it’s similar to what I experimented with a couple of years ago, but now it is supported by a complete infrastructure in the form of an MCP specification, libraries, and utilities. I really want to extend my sabbatical 🙂

At the same time, I tested how quickly one can create a website and all the surrounding infrastructure for a project. And it is amazing! In 3-4 hours, you can raise the perceived level of an OSS project from “someone hacked this together” to “it looks beautiful, there might be a team behind it.” And building small utilities like a configurator is a pure pleasure.