DeepCamera is an open-source AI camera skills platform and AI network video recorder that runs vision-language models locally on your own hardware. Instead of streaming your home to the cloud, it watches, understands, remembers, and guards it on-device, and talks to you through Telegram, Discord, or Slack.
DeepCamera is developed by SharpAI on GitHub, where it has earned more than 3,100 stars under the MIT license, with active development through September 2026. It performs local VLM video analysis using models such as Qwen, DeepSeek, SmolVLM, and LLaVA, alongside YOLO-based detection, so cameras can describe events in natural language rather than just flagging motion. Its architecture is built around pluggable AI skills: each capability, from person and face recognition to package detection, is a self-contained module with its own model and parameters, and a growing skill catalog spans security, home automation, and monitoring scenarios. The platform is LLM-agnostic, working with OpenAI, Google, or Anthropic APIs, or fully local models via llama.cpp, and it runs on modest hardware including Mac Minis, AI PCs, and Raspberry Pi boards. A desktop companion, SharpAI Aegis, adds LLM-powered setup, an agent chat interface for your cameras, and smart alerts connected to your phone, removing the need for manual Docker or command-line configuration. Home Assistant integration ties it into existing smart-home setups.
DeepCamera is free and open source under the MIT license. You supply the hardware and, if you choose them, any cloud model API keys; fully local operation costs nothing beyond electricity.
DeepCamera is for privacy-conscious homeowners, smart-home tinkerers, and small businesses that want AI-understood surveillance, who is at the door, what the dog did, when the package arrived, without handing continuous video of their lives to a cloud vendor.