Jetson Nano edge AI vision Agent Harness: real-time video analytics, MCP tools, event evidence, DeepSeek online/offline, secure recovery
🌐 中英双语项目主页 · ▶ 在线播放 8 段原始演示 · ⚡ 5 分钟快速验证 · 📦 Release 与 SBOM · 📚 文档导航
EdgeSentinel VisionOps 不是“摄像头接一个大模型”的一次性脚本。它把实时视觉、结构化事件、证据链、Agent 工具调用、MCP、权限策略、模型降级、可观测性与灾难恢复组织成一套可验证的边缘智能体工程。
项目在真实 Jetson Nano、USB 摄像头、Docker 与 systemd 环境中持续验收；在线模式连接 DeepSeek，网络或模型不可用时可明确降级到确定性离线规则，同时继续读取本地实时视觉结果。
[!NOTE] 本页媒体均采集自真实 Jetson Nano 实验环境。为完整呈现实验过程，演示素材保留原始画面、分辨率与音轨；生产部署应另行执行人物授权、标识脱敏和最小披露审查。
看得见 会行动 可治理 能恢复 --- --- --- --- Jetson 实时检测、跟踪、区域与库存 Agent 按需选择 Tool / Skill / MCP L0–L3、RBAC、确认门与脱敏 Trace systemd、熔断降级、加密备份与恢复演练
自研代码 核心应用 自动化测试 Agent / MCP 实机证明 ---: ---: ---: ---: ---: 约 86,800 行 / 365 文件 40,911 行 663 项通过 33 Tools · 25 MCP Tools · 6 Hooks 8 段原始视频
统计仅包含 Git 跟踪的自研源文件，排除 vendor、媒体、运行数据和发布产物；测试、运维代码与 13,347 行文档分别列示。详见工程规模与统计边界。
EdgeSentinel VisionOps is a self-hosted Jetson Nano edge AI video analytics and Agent Harness platform. It combines real-time object detection and tracking, people counting, zone events, inventory monitoring, tamper-evident event evidence, MCP tools, DeepSeek online/offline switching, policy-gated tool calling, observability, encrypted backup and disaster recovery in one verifiable computer-vision system.
- 为什么是 EdgeSentinel - English overview - 第一版工程规模 - 核心能力 - 硬件与器材 - 系统架构 - 安全与治理 - 实机证据与演示素材 - 快速验证 - 部署到 Jetson Nano - 可以怎样提问 - 项目结构 - 质量与可验证发布 - 文档导航 - 路线图 - 参与贡献
多数边缘视觉示例止步于“模型识别到了什么”。EdgeSentinel 继续回答四个工程问题：
1. 识别结果怎样成为可查询、可追溯的事件？ 2. Agent 调用设备工具时，怎样限制权限与副作用？ 3. 网络、模型、摄像头或服务异常时，系统怎样降级与恢复？ 4. 怎样证明一次回答、一次动作和一份发布包确实可信？
From the project README.
Add the radar badge to your README — it shows your project was picked up by MCP Radar and links to this page:
[](https://mcp.liqiwa.com/s/zja0011--edgesentinel-visionops.html)
dMoERA Creator Studio MCP Server: build, backtest, and deploy crypto trading strategies via AI agents
sergiorbk/petri-net-simAn editor, a simulator, and a natural-language generator for Petri nets and Petri-object models, in one place
fxd0h/Axelera-Voyager-Local-AssistantLocal AI assistant for the Axelera Voyager SDK and Metis AIPU — semantic RAG over the public Axelera sources + workflow skills, via MCP. No account, no cloud.
MuhittinYilmazer/akanaSelf-hosted, local-first AI assistant server: swappable LLM providers, review-gated memory, an encrypted vault, and wake-word voice. No accounts, no telemetry.
AdamRussak/self-docSelf-hosted documentation RAG pipeline that serves semantic search to LLM agents over the Model Context Protocol.
SujalXplores/Agent-KCode-enforced incident-response agent that investigates SigNoz alerts, publishes evidence-backed root-cause claims, and gates every action behind a safety policy. Built for Agents of SigNoz hackathon.
The top new MCP servers of the week, every Monday. No spam, unsubscribe anytime.