RSI Anything

RSI Anything is an open source framework that turns world-model predictions and interaction feedback into recursive self-improvement.

RSI-Anything is a domain-general framework that turns world-model predictions and interaction feedback into recursive self-improvement. Built on JEPA-Anything, it connects candidate generation, prediction-guided selection, external evaluation and learning in a shared process. The framework applies to code repair, GUI interaction, laboratory planning, reaction optimization and immune science. Each domain supplies its own state, actions, prediction targets and feedback while using the same cycle of proposing candidates, anticipating consequences, evaluating selected actions and retaining improvements. Domain adapters connect the shared workflow to each task's states, action space and external feedback.

RSI Anything at a glance

Best for: Software engineer, Researcher
Runs on: open source, Linux, CLI
Pricing: Free
Released: October 2026
Category: AI agent tool
Website: github.com/gen-verse/rsi-anything

Key features

Connect candidate generation and prediction-guided selection
Use external evaluation and learning in a shared process
Propose actions, predict consequences, and evaluate selections
Retain improvements for subsequent rounds
Support code repair, GUI interaction, and laboratory planning
Optimize reactions and analyze immune science workflows

Use it for

Rank candidate programs using predicted execution outcomes
Connect observations and action proposals for GUI tasks
Predict candidate plan quality for laboratory action sequences
Rank catalyst and solvent choices using yield predictions

Alternatives to RSI Anything

Rsi Eval — It supplies the evaluation boundary so methods can use existing benchmarks without reimplementing execution.
Reflexio — It autonomously extracts lessons from interactions to permanently correct agent behavior.
GEAR — Organizes recursive self-improvement through evolutionary search and reproducible evaluation.
OmniAgent — It evolves skills in real-time and dynamically hardens security during task execution.
Agent Cockpit — It checks evidence chains deterministically so broken claims never slip by unnoticed.
DiscoveryRSI — Achieves high actionable-reference rates and supports multi-agent evolutionary research search.
SoL Pi — Saves $8.75 to $13.50 per hour compared to native Codex and Claude Code systems.
Beeuni — Agents retain context and improve skills over time through a controlled review process.
II-Agent — It provides a fully open-source, transparent, and auditable agent foundation.
Harness2 — It recursively refines agent execution structures using parallel and sequential search.
Webwright — Enables code-based web automation with reusable and rerunnable scripts.
FastGPT — It uses hybrid retrieval and auto-cleansed data to prevent hallucinations.

All RSI Anything alternatives