AutoResearchClaw

What problem does it solve?

AutoResearchClaw automates the end-to-end process of turning a research idea into a conference-ready paper, removing manual, error-prone steps and enabling fully autonomous output from topic to citations and manuscript.

Core Features & Use Cases

• 23-stage autonomous pipeline spanning 8 phases (topic init to citation verification) with multi-agent orchestration
• Specialized agents (CodeAgent, BenchmarkAgent, FigureAgent) for code generation, benchmarking, and figure production
• Self-learning via MetaClaw cross-run lessons injected into prompts to improve future runs
• Hardware-aware execution that auto-detects CUDA/MPS/CPU and adapts code generation and experiment scale
• Real literature integration from OpenAlex, Semantic Scholar and arXiv with 4-layer citation verification
• OpenClaw-compatible bridge for natural language control and cross-tool automation
• Self-healing sandbox execution, robust QA gates, and reproducible artifact generation

Quick Start

Run the engine with a given topic to generate a full conference-ready paper autonomously.

Quick Start

Run AutoResearchClaw with a topic to generate a full paper autonomously.

Quick Start

Use the OpenClaw bridge to drive end-to-end research via natural language.

Quick Start

Initialize a run by configuring your LLM provider and topic, then execute the pipeline to obtain the paper and deliverables.

Quick Start

Start a new run with an interactive setup to tailor model selection and templates.

Quick Start

Start a run with a topic and an auto-approval flag to bypass gating.

Quick Start

Begin a full autonomous run and retrieve LaTeX, BibTeX, and figures