The rapid development of large language models (LLMs) has demonstrated great potential in automating document drafting. However, the preparation of tender documents involves a vast amount of domain-specific data, requiring not only structural completeness and precise clauses but also compliance with industry regulations and project requirements. Due to a lack of relevant domain knowledge, pretrained LLMs may generate tender documents with structural omissions, inaccuracies, or even fabricated clauses, making them unsuitable for practical applications. To address this, we propose TDAD, a tender document automated drafting framework with fine-tuned LLMs. We first construct an instruction dataset based on specific clause data and fine-tuned the LLM to enhance its expertise in the tendering domain. Then, we design targeted prompts to guide the model in generating comprehensive and compliant tender documents by integrating project requirements with the domain knowledge learned during fine-tuning. Experimental results show that this method effectively supports the automated drafting of tender documents, demonstrating strong performance in content completeness and accuracy, thus validating its feasibility and practical value.