Code comprehension remains a major bottleneck in maintaining large-scale legacy code repositories, particularly due to their complex intra-procedural control flows and lack of up-to-date documentation. While textual summaries generated by large language models (LLMs) provide semantic insights, they fail to intuitively represent non-linear control logic. Visual flowcharts are therefore highly desirable, yet existing automated approaches either produce cluttered diagrams or generate syntactically invalid scripts that cannot be rendered. To address these challenges, we present AutoFLC, a general-purpose tool to transform repository-scale legacy codebases into comprehensive visual flowcharts. AutoFLC combines static analysis for precise function extraction with a renderer-in-the-loop self-correction mechanism. This tool iteratively refines the generated PlantUML script to guarantee syntactic validity, thereby ensuring the successful rendering of the final visual flowchart. Evaluations on three real-world aerospace systems demonstrate that AutoFLC achieves a flowchart generation success rate exceeding 99%. A user study with 10 industrial practitioners further confirms that the generated flowcharts significantly reduce cognitive load and effectively support logic comprehension. The tool is publicly available at https://github.com/dfjdkjkdf/AutoFLC/tree/main under the MIT License, with a demonstration video accessible at https://youtu.be/AeEi9Nbw7xI.

@inproceedings{paperbot12355,
  title = {AutoFLC: An LLM-Based Flowchart Generation Tool for Legacy Code Repository},
  author = {Yixing Luo and Longxing Yang and Xiaofeng Li and Bin Gu and Zhi Jin},
  booktitle = {ASE},
  year = {2026},
  doi = {10.1145/3832783.3834635}
}