primeintellect-ai/prime-agent is an agent repository for coding, research, and long-running autonomous tasks. The latest release recorded on GitHub is v0.10.1, but the provided information contains no changelog for this version. The repository description introduces it as a self-improving RLM agent for coding tasks and long-running execution. The implementation language is Rust.
The core of this project is the recursive language model (recursive language model, RLM) architecture. It handles context like variables inside a persistent REPL and calls sub-agents like functions. A REPL is an execution environment where you enter commands and keep checking the results. According to the README, it preserves auxiliary prompts, memory, technical explanations, and reusable sub-agent settings as session state. On this basis, it is designed to carry on coding, research, and long-running tasks. The README also explains installation, background sessions, tool usage, and the structure of continuous execution.
However, the README explains that session state is, by default, limited to the current session. Therefore, there is no grounds for saying that state continues to be maintained across sessions. The license field is NOASSERTION, and no confirmed license name was provided. The changelog for v0.10.1 also cannot be confirmed, so it is hard to evaluate the improvements of that release from this material alone.