RLM CLI brings recursive reasoning to terminal
rlm-cli is an open-source terminal tool that applies Recursive Language Models by having the model write code to chunk, query, and aggregate long context instead of forcing everything into one prompt. It runs interactively or as a one-shot command, accepts files, directories, globs, URLs, or pasted text, supports Anthropic, OpenAI, Gemini, and OpenRouter, and saves trajectories for later inspection.
Hot take: this is a smart answer to context rot because it turns “reason over lots of data” into an explicit workflow instead of a single giant prompt.
- –The core differentiator is the recursive loop: chunk, query, aggregate, repeat until `FINAL()`.
- –It’s genuinely multi-provider, which lowers lock-in and makes it easier to fit into existing workflows.
- –Auto-loading project context from the current directory is the kind of ergonomics that can make this actually sticky for developers.
- –The saved trajectories are a strong product choice because they make the system debuggable, not just magical.
- –It’s early and solo-maintained, so the main risk is polish and reliability rather than the concept itself.
DISCOVERED
68d ago
2026-03-20
PUBLISHED
68d ago
2026-03-20
RELEVANCE
AUTHOR
Pinaka-X