TY - JOUR TI - annotaid : A browser-based tool for LLM-assisted qualitative coding of open text AU - Sprengholz, Philipp PY - 2026 JO - SoftwareX VL - 34 IS - 102702 SP - 1-4 PB - Elsevier CY - Amsterdam [u.a.] SN - 2352-7110 DO - 10.1016/j.softx.2026.102702 UR - https://fis.uni-bamberg.de/handle/uniba/115031 LA - eng AB - annotaid is a lightweight, browser-based tool that enables social science researchers to classify large collections of open text using locally-running large language models (LLMs). Operating entirely within the user's browser and communicating with a locally-running inference server, annotaid processes no data externally, making it suitable for research involving sensitive or confidential text such as interview transcripts or survey responses. The tool has native support for LM Studio and Ollama, and is compatible with any other inference backend that adheres to the OpenAI API protocol. Researchers define a coding scheme through a natural-language system prompt, optionally test it on individual items, and then apply it to an entire dataset provided as a CSV file. Results are returned as an enriched CSV with the model's output appended as a new column. The tool requires no programming knowledge and no cloud subscription, lowering the barrier to LLM-assisted qualitative coding for researchers without computational backgrounds. KW - Qualitative coding KW - Content analysis KW - Large language models KW - Local inference ER -