CompNav · Open-source projects
always accompany
A locally run AI companion app that combines character chat, memory, a desktop pet, and productivity modes.
What it is
always accompany is a locally run AI companion and personal assistant app. Character chat, roleplay, work or coding modes, and desktop presence share a memory and context system. Users import character cards, connect an AI provider, and access the local interface in a browser; optional capabilities include a Live2D or image-based desktop pet, screen awareness, and a Discord bot. Its scope is broad, and the project notes that companionship-oriented prompts are still being refined.
Core mechanisms and capabilities
Chat mode supports character cards and roleplay, while layered memory and retrieval bring prior information into new conversations. The desktop pet can use Live2D or image assets and supports screen awareness, proactive comments, and game companionship. Voice is currently described mainly as local speech-to-text; the README does not claim general speech synthesis.
Editorial take
It merits inclusion because it already provides character chat and desktop companionship rather than only an abstract agent framework. Its distinctive angle is carrying memory and character state across multiple modes alongside productivity tools. Users should note that the project also emphasizes productivity and says companionship-focused polish is still in progress; its broad feature set requires configuration.
Who it suits
Suited to users who want character chat, desktop companionship, and AI work tools in one place and are willing to configure a model API, character cards, and optional plugins.
Deployment and technical requirements
The project provides a Windows launcher and shell launchers for Linux or macOS; users access the local service in a browser. Hosted models require a personal API key, while application data is mainly local. Because it combines several workflows, first-run setup, plugins, and permissions are more involved than a single-purpose chat frontend.
Runtime requirements
First launch installs the Deno runtime and downloads dependencies. Users need a working AI API and must configure its endpoint, key, and model. Optional features such as the desktop pet, screen awareness, vector retrieval, and bots have separate settings and may add resource or privacy considerations.
The quick start requires a working AI API: enter its service URL, API key, and model, then choose or import a character card. The author says the local memory and retrieval service can be switched off. A local LLM is not required, but hosted-model cost and capability depend on the provider.
Limitations and considerations
The project says its companionship-oriented prompts and experience are still being refined, and the product also includes work and coding capabilities rather than being a dedicated virtual partner. When screen awareness or hosted models are enabled, interpreted screen context or chat may be sent to the selected AI service; review that provider's data practices.
Official sources and further reading
Information checked: Oct 26, 2026.