GhostWriter is an AI-powered writing environment that enhances collaborative human-AI writing experiences through personalization and agency. The system allows users to personalize the writing process by providing editable style and context information, and by explicitly teaching the system about their style preferences through highlighting likes and dislikes. GhostWriter leverages large language models (LLMs) to learn the user's intended writing style implicitly as they write, while allowing explicit teaching moments through manual style edits and annotations. A user study with 18 participants found that GhostWriter helps users craft personalized text generations and empowers them by providing multiple ways to control the system's writing style. The study revealed that users found GhostWriter useful for generating personalized text and championing user agency. The system offers a range of features, including the ability to refine existing text, generate new text, and teach the system about their target style through likes and dislikes. GhostWriter also allows users to define and refine context information. The system is designed to preserve user agency and ownership in the creative process, and to provide transparency into its internal state. The key contributions of this work include GhostWriter, an AI-enhanced writing environment that offers content personalization through style and context, user study results that reveal the utility of GhostWriter in generating personalized text generations and championing user agency, and a set of design recommendations and directions for future research on LLM-assisted writing. The system is built on the principles of leveraging machine capabilities while championing agency and control, using familiar editor metaphors, blending into the writer's existing workflow, and providing transparency to support reflection and discovery. The system's interface includes a main editor, left panel, and style toolbar, which allow users to interact with the system and customize their writing experience. The system is implemented as a web application with a Python backend that provides AI services through a RESTful endpoint. The system's AI services are orchestrated with the LangChain framework, and most operations use OpenAI's Chat GPT-4 model. The system's design is informed by related work in AI-assisted writing, personalized AI experiences, and working with style and context. The study found that participants found GhostWriter useful for generating personalized text and championing user agency, and that the system's features allowed users to interact with the system in various ways. The study also revealed that participants valued the system's transparency and the ability to provide feedback to the system. The system's design recommendations include providing multiple personalization paths, allowing users to define and select different styles and contexts, and encouraging granular feedback. The system's ability to provide personalized text generations and empower users through agency is a key contribution to the field of AI-assisted writing.GhostWriter is an AI-powered writing environment that enhances collaborative human-AI writing experiences through personalization and agency. The system allows users to personalize the writing process by providing editable style and context information, and by explicitly teaching the system about their style preferences through highlighting likes and dislikes. GhostWriter leverages large language models (LLMs) to learn the user's intended writing style implicitly as they write, while allowing explicit teaching moments through manual style edits and annotations. A user study with 18 participants found that GhostWriter helps users craft personalized text generations and empowers them by providing multiple ways to control the system's writing style. The study revealed that users found GhostWriter useful for generating personalized text and championing user agency. The system offers a range of features, including the ability to refine existing text, generate new text, and teach the system about their target style through likes and dislikes. GhostWriter also allows users to define and refine context information. The system is designed to preserve user agency and ownership in the creative process, and to provide transparency into its internal state. The key contributions of this work include GhostWriter, an AI-enhanced writing environment that offers content personalization through style and context, user study results that reveal the utility of GhostWriter in generating personalized text generations and championing user agency, and a set of design recommendations and directions for future research on LLM-assisted writing. The system is built on the principles of leveraging machine capabilities while championing agency and control, using familiar editor metaphors, blending into the writer's existing workflow, and providing transparency to support reflection and discovery. The system's interface includes a main editor, left panel, and style toolbar, which allow users to interact with the system and customize their writing experience. The system is implemented as a web application with a Python backend that provides AI services through a RESTful endpoint. The system's AI services are orchestrated with the LangChain framework, and most operations use OpenAI's Chat GPT-4 model. The system's design is informed by related work in AI-assisted writing, personalized AI experiences, and working with style and context. The study found that participants found GhostWriter useful for generating personalized text and championing user agency, and that the system's features allowed users to interact with the system in various ways. The study also revealed that participants valued the system's transparency and the ability to provide feedback to the system. The system's design recommendations include providing multiple personalization paths, allowing users to define and select different styles and contexts, and encouraging granular feedback. The system's ability to provide personalized text generations and empower users through agency is a key contribution to the field of AI-assisted writing.