MAGIC is a novel multi-agent method for generating self-correction guidelines for text-to-SQL. It automates the creation of guidelines that help large language models (LLMs) revise their previously incorrect SQL queries. MAGIC consists of three agents: a manager, a feedback agent, and a correction agent. These agents collaborate to iteratively generate and refine a self-correction guideline tailored to LLM mistakes. The method outperforms human-created guidelines in experiments, improving the interpretability of corrections and providing insights into LLM performance in self-correction. MAGIC's guidelines are publicly available to foster further research in automatic self-correction guideline generation. The method is designed to address the limitations of existing self-correction approaches, which rely on human expertise and are labor-intensive. MAGIC uses a multi-agent framework to analyze the failures of an initial text-to-SQL method and automatically generate a self-correction guideline tailored to those mistakes. The manager agent iteratively interacts with the feedback and correction agents to refine the guideline. The feedback agent provides explanations of mistakes, and the correction agent revises the SQL queries based on this feedback. The manager agent then stores successful feedback and aggregates it to generate the self-correction guideline. Experiments show that MAGIC's guidelines significantly improve the effectiveness of text-to-SQL methods. MAGIC's guidelines outperform human-created guidelines in execution accuracy and self-correction performance across different datasets and scenarios. The method is efficient, with guidelines generated in under two hours. MAGIC's guidelines are also adaptable to different text-to-SQL systems and can be applied to various databases. The study also highlights the importance of self-correction in text-to-SQL, where LLMs are prone to errors. MAGIC's approach provides a systematic way to generate self-correction guidelines that can be used to improve the accuracy and efficiency of SQL queries. The method is applicable to other tasks and can be used to enhance the performance of LLMs in various domains. The findings contribute to advancing the state-of-the-art in text-to-SQL translation and provide valuable insights for future research in this area.MAGIC is a novel multi-agent method for generating self-correction guidelines for text-to-SQL. It automates the creation of guidelines that help large language models (LLMs) revise their previously incorrect SQL queries. MAGIC consists of three agents: a manager, a feedback agent, and a correction agent. These agents collaborate to iteratively generate and refine a self-correction guideline tailored to LLM mistakes. The method outperforms human-created guidelines in experiments, improving the interpretability of corrections and providing insights into LLM performance in self-correction. MAGIC's guidelines are publicly available to foster further research in automatic self-correction guideline generation. The method is designed to address the limitations of existing self-correction approaches, which rely on human expertise and are labor-intensive. MAGIC uses a multi-agent framework to analyze the failures of an initial text-to-SQL method and automatically generate a self-correction guideline tailored to those mistakes. The manager agent iteratively interacts with the feedback and correction agents to refine the guideline. The feedback agent provides explanations of mistakes, and the correction agent revises the SQL queries based on this feedback. The manager agent then stores successful feedback and aggregates it to generate the self-correction guideline. Experiments show that MAGIC's guidelines significantly improve the effectiveness of text-to-SQL methods. MAGIC's guidelines outperform human-created guidelines in execution accuracy and self-correction performance across different datasets and scenarios. The method is efficient, with guidelines generated in under two hours. MAGIC's guidelines are also adaptable to different text-to-SQL systems and can be applied to various databases. The study also highlights the importance of self-correction in text-to-SQL, where LLMs are prone to errors. MAGIC's approach provides a systematic way to generate self-correction guidelines that can be used to improve the accuracy and efficiency of SQL queries. The method is applicable to other tasks and can be used to enhance the performance of LLMs in various domains. The findings contribute to advancing the state-of-the-art in text-to-SQL translation and provide valuable insights for future research in this area.