mediar-ai/screenpipe
Free本地优先系统，捕获屏幕/音频并带时间戳索引，SQL/嵌入存储，语义搜索，LLM 历史分析，事件触发动作。通过 NextJS 插件生态系统构建上下文感知 AI 代理。
About mediar-ai/screenpipe
Screenpipe is a source-available application that continuously captures your screen and audio 24/7, storing all data locally on your device. It indexes everything you see, say, and do with timestamps, enabling AI-powered semantic search, natural language queries, and automated workflows. Features include full accessibility tree capture, OCR, speech transcription, speaker identification, keyboard inputs, and app switches. It consumes 5-10% CPU, 0.5-3 GB RAM, and about 20 GB of storage per month. Filters allow selective recording by window, app, or content, and an optional proprietary AI PII model runs in 9ms on consumer hardware. The data can be accessed via SQL, embeddings, or APIs, and integrated with AI agents (e.g., Claude, OpenClaw) through a plugin ecosystem called Pipes. Screenpipe works offline, supports end-to-end encryption at rest, and is built to be private and transparent—users can inspect, modify, and audit the code.
Key Features
Pros & Cons
- 100% local processing ensures complete privacy and data control
- Works offline with no internet dependency
- Low resource footprint relative to cloud-based alternatives
- Source-available code allows auditing, customization, and community contributions
- AI search and automation without uploading data to third parties
- Free to use with all features included
- Supports encryption at rest for added security
- Active development with frequent updates and active community (Discord, GitHub)
- Continuous recording consumes ~20 GB of storage per month
- Uses 5-10% CPU and 0.5-3 GB RAM even when idle
- May raise privacy concerns for shared or corporate devices despite local storage
- Source-available license may not meet all enterprise open-source requirements
- Requires constant running background process, impacting battery on laptops