# VideoDB > VideoDB is real-time video infrastructure for AI agents — it gives agents eyes, ears, and memory. It turns video files, YouTube URLs, live RTSP camera streams, meetings, and screen recordings into structured, searchable context that AI applications and agents can retrieve, subscribe to, and act on. VideoDB exposes its entire marketing site as Markdown alongside the HTML. Every page has a `.md` counterpart at the same path (e.g. `/platform` ↔ `/platform.md`). Each HTML page also declares this with `` in its ``. The full site is also available concatenated at `/llms-full.txt`. ## Key facts - What it does: ingest any video source → continuously understand it (speech + visual analyzers) → index it → semantic search with timestamped results → compile evidence clips → real-time events and alerts (WebSocket/webhooks) defined in plain language (e.g. "detect when a person appears"). - Real-time: RTSP Connect ingests live camera, drone, and broadcast streams; understanding runs in rolling time windows (e.g. every 5s); live indexes are queryable while the stream runs. - YouTube & files: `coll.upload(url="https://www.youtube.com/watch?v=...")` ingests YouTube URLs directly; agents can watch, search, and cite videos with playable timestamped clips. - Screen capture SDK: gives computer-use and browser-use agents persistent, searchable visual memory of desktop and browser sessions. - Model-agnostic: VLM/ASR analyzers are configurable by prompt; partner models (e.g. TwelveLabs Pegasus) can run inside VideoDB indexing pipelines. - Install: `pip install videodb` (Python) · `npm install videodb` (Node.js/TypeScript) · MCP server: `uvx videodb-director-mcp` · Agent skills: `npx skills add video-db/skills`. - Integrations: MCP clients (Claude, Cursor), LlamaIndex retriever (`llama-index-retrievers-videodb`), LangChain, n8n, Zapier. - Enterprise: SOC 2 Type II, GDPR, HIPAA, ISO 27001; Zero Data Retention options; SAML/OIDC SSO, SCIM, RBAC; managed cloud or bring-your-own-cloud (AWS/GCP/Azure), VPC, edge GPU. - Not video generation: VideoDB is video understanding/perception infrastructure (analysis, search, memory, events), not an AI video generator or avatar tool. - Free tier at https://console.videodb.io — first semantic search in about 5 minutes. ## Core - [Homepage](https://videodb.io/index.md): What VideoDB is and who it's for. - [Platform](https://videodb.io/platform.md): One backend, six primitives — ingest, index, search, edit, stream, deploy. - [Developers](https://videodb.io/developers.md): SDKs, quickstarts, examples, and reference. - [Pricing](https://videodb.io/pricing.md): Plans from solo developers to enterprise — transparent unit rates for ingest, indexing, search, streaming, and generation. - [Company](https://videodb.io/company.md): About, team, partners, investors, contact. ## Solutions - [Agentic Perception](https://videodb.io/agentic-perception.md): Eyes, ears, memory for agents. - [Agents on the Desktop](https://videodb.io/agentic-perception/desktop.md) - [Agents in Sandboxes](https://videodb.io/agentic-perception/sandboxes.md) - [Agents on the Web](https://videodb.io/agentic-perception/web.md) - [Live Camera Intelligence](https://videodb.io/live-camera-intelligence.md): Real-time monitoring for cameras and sensors. - [Drone Operations](https://videodb.io/live-camera-intelligence/drone.md) - [Home Security](https://videodb.io/live-camera-intelligence/home-security.md) - [Tele-ICU](https://videodb.io/live-camera-intelligence/icu.md) - [Programmable Media](https://videodb.io/programmable-media.md): Query, edit, stream by code. - [Case Study: OTT](https://videodb.io/programmable-media/case-study-ott.md) - [Case Study: Brand Marketing](https://videodb.io/programmable-media/case-study-brand-marketing.md) - [Training Video Data](https://videodb.io/training-video-data.md): Training data infrastructure for world models, VLAs, and physical AI. - [World Model Data](https://videodb.io/world-model-data.md): Training data for physical AI and world models. ## Platform deep-dives - [Personalized Streaming](https://videodb.io/platform/personalized-streaming.md): Per-viewer streaming with personalization, ads, and edits computed in real time. - [Realtime Alert](https://videodb.io/platform/realtime-alert.md): Event-driven alerting on live video. ## Research - [Research index](https://videodb.io/research.md): Papers, evaluations, and technical reports. - [Search over the Visual World](https://videodb.io/research/search-over-the-visual-world.md): Technical report — persistent visual memory, layered indexes, source-grounded evidence; 73.09% macro Recall@1 over 9,834 queries across 4 public benchmarks vs 65.75% for a commercial video-native engine ([PDF](https://labs.videodb.io/papers/search-over-the-visual-world.pdf)). ## Blog - [Blog index](https://videodb.io/blogs.md): Product updates, partnerships, engineering deep dives, and tutorials. - [Search over the Visual World](https://videodb.io/blogs/search-over-the-visual-world.md): Full VideoDB technical report (July 2026) — visual memory, layered indexes, source-grounded evidence, and a 9,834-query retrieval evaluation. - [Blog index](https://videodb.io/blogs.md): Guides, tutorials, comparisons, and essays on giving AI agents eyes. - [Claude + VideoDB Skills Edited Our Launch Video](https://videodb.io/blogs/claude-edited-our-launch-video.md): An agent indexed the raw takes, picked the best take per beat, rendered the cut, then re-indexed its own output to audit it. - [Give Your AI Agents Eyes](https://videodb.io/blogs/give-your-ai-agents-eyes.md): The builder's guide to real-time visual perception — YouTube, live cameras, and screens, with working code. - [Your AI Agent Can't Watch YouTube. Here's the Fix.](https://videodb.io/blogs/ai-agent-watch-youtube.md): YouTube videos as searchable, timestamped agent context in ~20 lines of Python. - [RTSP + AI: Turn Any Camera Stream into Agent-Readable Events](https://videodb.io/blogs/rtsp-ai-analysis.md): Continuous understanding, plain-language alerts, searchable history for any RTSP feed. - [Video RAG: The Definitive Guide (2026)](https://videodb.io/blogs/video-rag.md): The four architectures, working code, and RAG over live streams. - [Twelve Labs Alternatives in 2026](https://videodb.io/blogs/twelve-labs-alternatives.md): An honest, job-to-be-done guide to the video AI landscape. - [What Video Retrieval Benchmarks Get Wrong About Ground Truth](https://videodb.io/blogs/video-benchmark-ground-truth-ambiguity.md): Manual review of 5 canonical benchmarks with timestamped clip evidence. - [JEPA: From Language Models to World Models](https://videodb.io/blogs/jepa-from-language-models-to-world-models.md): What JEPA is and why the training target moves from tokens to hidden state. - [How to Evaluate Multimodal VLMs for Your Video Use Case](https://videodb.io/blogs/how-to-evaluate-multimodal-vlms-for-your-video-use-case.md): Practical eval workflow with an open-source harness. - [Strong VLMs Can Still Fail on Downstream Vision Tasks](https://videodb.io/blogs/claude-chessboard-spatial-reasoning.md): 72-position chessboard-to-FEN eval of frontier VLMs. - [VideoDB × TwelveLabs](https://videodb.io/blogs/twelvelabs.md): Search any moment across your video library. - [VideoDB × LlamaIndex](https://videodb.io/blogs/llama-index.md): Plug video into your RAG pipeline. - [TinyFish × VideoDB](https://videodb.io/blogs/tinyfish.md): TinyFish opens the web for agents, while VideoDB lets them see inside video and turn access into comprehension. - [Why AI Agents Are Blind Today](https://videodb.io/blogs/why-agents-are-blind.md): The gap between human perception and agent perception, and why it matters for AI. - [Perception Is the Missing Layer](https://videodb.io/blogs/perception-is-the-missing-layer.md): Why LLMs and RAG need perception of the world as it happens. - [MP4 Is the Wrong Primitive for AI](https://videodb.io/blogs/mp4-is-wrong-primitive.md): Why AI agents need video indexes, not opaque playback files. - [Why Video Was Built for Playback, Not Perception](https://videodb.io/blogs/playback-vs-perception.md): How video infrastructure optimized for human playback leaves AI perception behind. - [What Episodic Memory Means for AI Agents](https://videodb.io/blogs/episodic-memory-for-agents.md): Why agents need verifiable, playable memory of experiences. - [Infrastructure that Sees and Edits](https://videodb.io/blogs/infrastructure-that-sees-and-edits.md): VideoDB infrastructure for combining visual understanding with programmable video editing. ## Trust & legal - [Trust Centre](https://videodb.io/trust-centre.md): Compliance, certifications, and security posture — SOC 2 Type II, ISO 27001, HIPAA, GDPR. - [Terms of Service](https://videodb.io/legal/terms.md): Terms and conditions for VideoDB. - [Privacy Policy](https://videodb.io/legal/privacy.md): How we collect, use, and protect personal data. - [Security & Compliance](https://videodb.io/legal/security.md): Encryption, access control, secure data centres. - [DPA](https://videodb.io/legal/dpa.md): Standard Data Processing Agreement. ## External - [Documentation](https://docs.videodb.io): SDKs, API reference, guides — with its own [llms.txt](https://docs.videodb.io/llms.txt). - [Quickstart](https://docs.videodb.io/pages/getting-started/quickstart): First upload → understand → index → search in ~5 minutes. - [MCP Server](https://docs.videodb.io/pages/build-with-agents/mcp-server): Video perception tools for Claude, Cursor, and any MCP client. - [GitHub](https://github.com/video-db): Source for SDKs, Director, the cookbook, and agent-toolkit. - [PyPI](https://pypi.org/project/videodb/) · [npm](https://www.npmjs.com/package/videodb) - [Discord](https://discord.com/invite/py9P639jGz) - [Status](https://status.videodb.io)