# Twelve Labs MCP for AI Agents AI Agent Connect > Twelve Labs MCP for AI Agents. It lets you search, analyze, and extract insights from video content using natural language. You can index large video libraries, find specific scenes, and pull structured data from visual and audio information. It turns video archives into searchable data for content teams and security operations. ## Overview - **Category:** ai-frontier - **Price:** Free - **Endpoint:** https://edge.vinkius.com/vk_preview_Vgl5VVTXIogBXlRhR1Xxpl2YsHWFIR0xQ7beKVtS/ai-agent-connect - **Tags:** video-search, multimodal-ai, semantic-search, video-indexing, computer-vision ## Description This Connector lets you treat your video library like a searchable database. Instead of scrolling through hours of footage or manually tagging every clip, you can just ask your AI agent to find what you need. If you're a video editor, you can search for a specific b-roll shot or a scene where a specific action happens using plain English. The Connector handles the heavy lifting of indexing and semantic understanding, turning raw video into actionable data. You can upload files via URLs or multipart sessions and keep track of their progress as they get added to your library. It also lets you generate embeddings and manage collections of entities, like specific people or objects, which is a game changer for building custom machine learning workflows. By connecting this through the Vinkius catalog, you give your agent the ability to watch and interpret your archives in seconds. You get structured insights and exact timestamps instead of having to hunt for them yourself. It removes the bottleneck of manual video management and lets you focus on the creative or analytical work that actually matters. ## Tools ### embed_async Generate multimodal embeddings for a batch of videos in the background. This is great for building large-scale search databases. ### analyze_async Run video analysis in the background. This is useful for processing multiple files at once without waiting. ### analyze_sync Perform immediate video analysis for quick results. This works well when you need a fast answer on a single clip. ### confirm_multipart_upload Finalize a multi-part upload session. This tells the system the file is ready for indexing. ### create_asset Upload a video file to the platform. This creates the base asset for your AI to analyze later. ### create_entity Add a specific person or object to a collection. This allows for granular tracking of specific subjects within your videos. ### create_index Set up a new video index with specific settings. This organizes your footage into searchable categories. ### create_multipart_upload Start a session to upload large video files in pieces. This helps keep your uploads stable for big files. ### delete_index Remove a video index from your account. Use this to clean up old or unnecessary archives. ### get_index Fetch details for a specific video index. Use this to check the status or name of an existing library. ### get_indexed_asset Retrieve information about a specific indexed video. This helps your agent find the right file from a large library. ### update_index Change the name of an existing video index. Use this to keep your organized libraries clearly labeled. ### create_entity_collection Group specific objects or people into a custom collection. This helps the AI remember specific recurring subjects. ### embed_sync Create embeddings for a video immediately. Use this for quick, real-time data generation on single clips. ### index_asset Add a newly uploaded video to an existing index. This makes the content searchable by your agent. ### list_indexes See all the video indexes you have created. This helps you navigate your organized video libraries. ### report_multipart_progress Check how much of a large video upload has finished. This keeps your agent informed on the status of large files. ### search Look for specific moments or actions in your videos. This is the primary way your agent finds content using natural language. ## Prompt Examples **Prompt:** ``` I have a huge folder of security footage. Can you find every time someone enters the back door? ``` **Response:** ``` I can certainly help with that. Please provide the index ID for your security footage. Once I have that, I'll scan the library for any instances of people entering the back door and provide a list of timestamps and clips. ``` **Prompt:** ``` Can you find any clips of people presenting a slideshow in my 'Webinar-Archive' index? ``` **Response:** ``` Searching the 'Webinar-Archive' index now... I found 3 matches: - **04:12** in 'Tech_Talk_01.mp4' - **12:45** in 'Product_Demo_A.mp4' - **33:20** in 'Live_Stream_04.mp4' Would you like me to summarize the content of these clips? ``` **Prompt:** ``` I need to create a new index for my social media ads. Can you set that up? ``` **Response:** ``` I can set that up for you. What would you like to name the index? Also, let me know if you want to include specific visual or audio analysis settings for the videos you add to it. ``` ## Capabilities ### Find specific scenes using natural language Ask your agent to locate specific moments in your video library using plain English descriptions. ### Index large video libraries for fast retrieval Organize your entire video archive into searchable indexes for rapid access. ### Upload and track video assets via URLs Add new videos to your library via direct links or multipart uploads and monitor their status. ### Extract structured data from video files Turn visual and audio content into structured data points using synchronous or asynchronous analysis. ### Generate multimodal embeddings for ML workflows Create embeddings for your video assets to support advanced machine learning and search tasks. ### Manage collections of specific entities Create and manage groups of specific people, objects, or items within your video content. ## Use Cases ### Finding b-roll A video editor asks the agent to find "a sunset over a mountain" in a 500-hour archive. The agent uses `search` to return three exact timestamps. ### Security review A security lead asks "show me every time a person enters the back door." The agent uses `search` across a specific index to list the moments. ### Content tagging A media team wants to extract all product names from a series of ads. The agent runs `analyze_sync` to pull out the data into a table. ### ML training A developer needs embeddings for a video dataset. The agent uses `embed_async` to process a batch of 100 videos for a custom model. ## Benefits - Stop scrolling through timelines by using `search` to find exact moments in seconds. - Organize massive archives into logical categories using `create_index` for faster retrieval. - Get structured data from raw footage using `analyze_sync` or `analyze_async` without manual tagging. - Manage large files reliably by using `create_multipart_upload` to ensure data integrity. - Build custom machine learning workflows by generating multimodal data with `embed_sync` and `embed_async`. - Track specific people or objects across your library by using `create_entity` and `create_entity_collection`. ## How It Works The bottom line is your AI agent gets the ability to query and analyze video content as easily as it reads text. 1. Connect your Twelve Labs API key to the Connector. 2. Upload your videos or provide URLs for indexing. 3. Ask your AI client to search for specific scenes or actions. ## Frequently Asked Questions **Can I use the Twelve Labs MCP to search my own private videos?** Yes. You can connect your private video library by creating indexes and uploading your files. Once indexed, your AI agent can search through them using natural language. **How does the Twelve Labs MCP handle large video files?** It supports multipart uploads, which allows you to break down and upload large files in pieces. This ensures that your uploads stay stable and complete even for very large assets. **Can the Twelve Labs MCP identify specific people in my footage?** Yes, it can. You can create entity collections to track and identify specific people or objects across your entire video library. **Does the Twelve Labs MCP work with my existing video library?** It works with any video content you can provide via URL or direct upload. You just need to create an index and add your assets to it so the agent can see them. **How do I get structured data out of my videos using the Twelve Labs MCP?** You can use the analysis tools to extract structured data from your videos. Your agent can then present this data in tables, summaries, or other formats you need. **Can I use the Twelve Labs MCP to find specific scenes by typing a description?** That is exactly what it is built for. You can type descriptions like "a person running in the park" and the Connector will find the relevant moments for you. **How do I list all my existing video indexes?** You can use the `list_indexes` tool. It will return a list of all indexes available in your Twelve Labs account, including their IDs and configuration. **Can I search for a specific moment inside my videos using text?** Yes! Use the `search` tool by providing an `index_id` and a search query. The AI will find the most relevant timestamps and video segments based on your description. **How do I add a new video to an index for analysis?** First, use `create_asset` with a public URL to upload the video. Then, use the `index_asset` tool with the resulting `asset_id` and your target `index_id` to start the processing.