LinkDigest
LinkDigest is a SaaS platform that converts social media URLs into structured text that AI agents can read. When a model is given a link from Xiaohongshu, Douyin, YouTube, TikTok, X, or standard web pages, a simple HTTP request returns a digest containing the post caption, transcript, OCR-extracted on-screen text, image descriptions, metadata, and key points in Markdown or JSON format. The service is specifically designed for LLM pipelines and MCP-compatible coding tools, allowing agents to automatically fetch social content that would otherwise render as JavaScript shells or login walls. LinkDigest works in three stages: resolve, read, and return. It expands short links and removes tracking parameters so repeated requests can be served from a shared cache. For reading, it uses platform-published captions when available; otherwise it transcribes audio and samples video frames to extract burned-in text. It never stores the source media, only the text digest. The digest includes timestamped transcripts, fragments of on-screen text, image descriptions, key points, and a pre-rendered Markdown version that can be inserted directly into a prompt. Pricing starts with three free digest requests without requiring a credit card, after which a paid Dev plan is $9 per month for 500 digests. Cached digests are shared between users, which means popular links are instant and do not count against quotas. LinkDigest exposes both a REST API and an MCP server for integration with tools like Claude Code and Cursor, and its documentation provides examples for social platforms that are typically difficult to scrape. Current platform coverage includes Xiaohongshu (RedNote), Douyin, YouTube, TikTok, X (Twitter), and general web pages. The team maintains an honest status table and reports when platforms block the service, such as Bilibili, instead of returning fabricated content. Some platforms are planned, like Instagram. LinkDigest advertises strong Mandarin speech recognition using Sense
- Category
- Analytics
- Pricing
- Paid
Problem solved
AI agents frequently fail to read social media links because fetching the URL returns only a JavaScript shell, a login wall, or app-download prompts. LinkDigest solves this by extracting the content into LLM-ready text, including transcripts, OCR text, and image descriptions, via one API call.
Use cases
- Provide a Xiaohongshu post as context to an AI coding agent in Claude Code or Cursor
- Extract YouTube video transcripts for summarization or content analysis
- Turn Douyin videos or image posts into structured text for search and indexing
- Read X posts with images or video as part of an agent's web research
- Efficiently process shared links through cached digests to avoid repeated API costs
- Set up an MCP server to let agents automatically digest any social link they encounter
- Retrieve image descriptions and OCR text from image-heavy social media posts
- Build LLM pipelines that monitor social links for mentions or qualitative data
Audience
- Developers
- ML Engineers
- Data Scientists
- Researchers
Pros
- Simple REST API and MCP server for quick integration
- Handles platforms that are otherwise inaccessible (Xiaohongshu, Douyin)
- Cached digests are shared and free for repeats, reducing cost
- Returns rich, structured Markdown/JSON with metadata and key points
- Privacy-conscious: does not store source media, only text digests
Cons
- Supports only a limited set of platforms; bilibili/Instagram are blocked or in progress
- Paid subscription required after 3 free digests ($9/month for 500 digests)
- Platform coverage can break due to third-party blocking
- Requires sending social media links to a third-party service
- Transcription/OCR may have errors, especially for uncommon languages
Official website: https://linkdigest.dev/
Canonical LambdaBase page: https://www.lambdabase.com/tools/linkdigest