Tenure

Tenure is a local, privacy-first proxy designed to provide long-term memory for Large Language Models (LLMs). It acts as an intermediary between any OpenAI-compatible client (like Open WebUI or LM Studio) and the LLM, automatically injecting relevant context based on a structured "world model" of user preferences, decisions, entities, and expertise. This eliminates the need to repeatedly brief the LLM on past conversations or specific requirements, ensuring that every new session is already contextualized. The tool is ideal for developers, writers, researchers, or anyone who frequently interacts with LLMs and experiences the frustration of models forgetting previous context. By running entirely on the user's machine, Tenure ensures data privacy and keeps all memory local. It supports instant import of existing knowledge and offers full control over the stored beliefs, allowing users to edit, audit, and manage their personalized context. Tenure aims to make LLM interactions more efficient and personalized by maintaining a persistent, structured understanding of the user's work. Tenure works by routing any OpenAI-compatible client to a local address (localhost:5757/v1). It intercepts prompts, enriches them with relevant information from its world model, and then forwards them to the chosen LLM provider (which can also be local or cloud-based). The LLM client remains unaware of Tenure's presence, making it a drop-in solution without requiring custom integrations or plugins. It also includes features for automatic history and belief compaction, and the ability to pause context extraction.