PhialsBasement/Chain-of-Recursive-Thoughts

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

2328 stars · 122 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

CoRT (Chain of Recursive Thoughts) is a tool that makes LLMs recursively refine their answers by generating multiple alternative responses, evaluating them, and picking the best one over several thinking rounds. It ships as a Python CLI script plus an early-stage web UI, and works with models via OpenRouter.

Use cases

• make a small local LLM reason better on coding tasks
• improve AI response quality by having the model argue with itself
• run recursive self-evaluation loops over LLM answers
• compare and pick the best of several AI-generated responses
• boost Mistral or other OpenRouter models' reasoning depth
• experiment with iterative LLM refinement techniques

When to choose

• you want better reasoning from small local models without fine-tuning
• you use OpenRouter or compatible APIs and want a drop-in thinking loop
• you want a simple script or early web UI to experiment with self-critique prompting

When to avoid

• you need production-grade, latency-sensitive inference (multiple rounds multiply cost and time)
• you need a maintained library with a stable API rather than an experimental project
• you want built-in support for providers other than OpenRouter without code changes

Facets

application · maturity active

llm-inference agent-framework prompt-engineering large-language-models artificial-intelligence developer-tools python cross-platform cli recursive-thinking self-evaluation reasoning-enhancement openrouter llm-wrapper web-ui nodejs

Member repositories

For agents

markdown · JSON · MCP: product_card(name="PhialsBasement/Chain-of-Recursive-Thoughts")