FARSIGHT is an innovative in-situ interactive tool designed to help AI prototypers identify potential harms associated with their large language model (LLM)-powered applications during the early prototyping stage. The tool addresses the challenge of anticipating potential risks and harms, which is particularly difficult for those without extensive AI expertise. FARSIGHT integrates with prompt-based AI prototyping tools, providing an always-on alert symbol that indicates the approximate alert level of a user's prompt. When the alert symbol is clicked, an Awareness Sidebar expands to highlight relevant news articles and LLM-generated potential use cases and harms. Users can interactively explore and edit these elements using the Harm Envisioner, a node-link diagram that visualizes potential consequences, stakeholders, and harms. The tool is designed to be model-agnostic and environment-agnostic, making it accessible to a wide range of users. The evaluation study with 42 AI prototypers found that FARSIGHT significantly improved users' ability to identify potential harms and was more useful and usable than existing resources. Participants also reported that FARSIGHT encouraged them to focus on end-users and consider indirect stakeholders and cascading harms. The tool is publicly available and open-source, aiming to facilitate the adoption of responsible AI practices in AI prototyping.FARSIGHT is an innovative in-situ interactive tool designed to help AI prototypers identify potential harms associated with their large language model (LLM)-powered applications during the early prototyping stage. The tool addresses the challenge of anticipating potential risks and harms, which is particularly difficult for those without extensive AI expertise. FARSIGHT integrates with prompt-based AI prototyping tools, providing an always-on alert symbol that indicates the approximate alert level of a user's prompt. When the alert symbol is clicked, an Awareness Sidebar expands to highlight relevant news articles and LLM-generated potential use cases and harms. Users can interactively explore and edit these elements using the Harm Envisioner, a node-link diagram that visualizes potential consequences, stakeholders, and harms. The tool is designed to be model-agnostic and environment-agnostic, making it accessible to a wide range of users. The evaluation study with 42 AI prototypers found that FARSIGHT significantly improved users' ability to identify potential harms and was more useful and usable than existing resources. Participants also reported that FARSIGHT encouraged them to focus on end-users and consider indirect stakeholders and cascading harms. The tool is publicly available and open-source, aiming to facilitate the adoption of responsible AI practices in AI prototyping.