Description
VidCrawl provides a specialized Video Context API that enables AI agents to efficiently process and understand video content from major platforms such as YouTube (including Shorts), TikTok, Instagram Reels, and X. The API transforms videos into structured data, including timestamped transcripts, visual storyboards, and high-resolution frames, delivered in both Markdown and JSON formats.
Read More
The core functionality revolves around a layered approach to video analysis. Instead of sending an entire video to a multimodal model, which can be slow and costly in terms of tokens, VidCrawl allows agents to start with text and small storyboard images. Agents can then request sharp frames only when specific visual details are needed, optimizing speed and cost. The API offers four main endpoints: `/v1/scrape` for comprehensive text data (transcripts, chapters, metadata in Markdown), `/v1/frames` for sharp, full-resolution images at precise timestamps, `/v1/search` for finding videos, and `/v1/agent` which automates the reading, glancing, and zooming process to answer questions with cited timestamps.
VidCrawl ensures deterministic output for `/v1/scrape` and `/v1/frames` endpoints, meaning the same video will always return the same text and frames without relying on LLMs for these specific operations. Only the `/v1/agent` endpoint utilizes a model, giving users control over when to engage it. The service supports public videos and provides clear error messages for private, age-restricted, or deleted content. If a video lacks captions, VidCrawl transcribes the speech with timestamps, indicating coverage for longer videos.
Target users include research agents needing to quickly digest long-form content, coding assistants requiring specific visual references from tutorials, and content pipelines for generating show notes, newsletters, or RAG data. The API is designed for easy integration, described in `/llms.txt` and an OpenAPI 3.1 specification, allowing coding agents to interact without custom SDKs. VidCrawl operates on a credit-based system, with 100 free credits provided upon account creation and one-time credit packs available for purchase, which never expire.
- CategoryNatural Language Processing Tools
- Pricing ModelFreemium
- DeploymentCloud
- PlatformsWeb
VidCrawl's Core Features
Video Context API for AI agents
Converts YouTube, TikTok, Instagram Reels, and X videos
Generates timestamped transcripts
Creates visual storyboards
Extracts sharp, full-resolution frames
Outputs data in Markdown and JSON formats
Deterministic output for transcripts and frames
Credit-based pricing model
100 free credits on signup
Supports public videos only
Transcribes speech for videos without captions
Optimized for cost and speed with layered data access
OpenAPI 3.1 specification for agent discovery
API key authentication for requests
Playground available for free video reading
How to use VidCrawl?
Sign Up: Create a free account to receive 100 credits and access full API features.
Obtain API Key: Your API key is provided upon account creation for authentication.
Select Endpoint: Choose the appropriate API endpoint based on your needs (e.g., /v1/scrape for transcripts, /v1/frames for images).
Send Video URL: POST a public video URL to the chosen endpoint with your API key.
Receive Structured Data: Get timestamped transcripts, storyboards, or sharp frames in Markdown or JSON.
Utilize Agent Endpoint: For complex queries, use the /v1/agent endpoint to automatically process video content and get answers with citations.
VidCrawl's Use Cases
- Research Agent Context
- Coding Assistant Guidance
- Content Pipeline Automation
- Brand Monitoring & Search
- Video Content Analysis
- AI Agent Development








