Supadata

What Supadata MCP server does

The Supadata MCP server exposes Supadata’s video and web data operations through the Model Context Protocol. It belongs in the Search & Data Extraction category and can provide transcript text, media metadata, webpage content, discovered URLs, multi-page crawl results, or structured fields extracted from video content.

Video support covers YouTube, TikTok, Instagram, Twitter, and file URLs for transcript extraction. Media metadata requests support YouTube, TikTok, Instagram, and Twitter URLs. For websites, the server separates single-page scraping from URL mapping and multi-page crawling, so an agent can first identify pages and then collect content from selected or related URLs.

How it works

The Supadata MCP server offers MCP tools that clients invoke with a URL and, where applicable, language, limits, prompts, or schemas. Transcript and scrape operations return content directly in text, Markdown, or HTML-oriented formats. Metadata and structured extraction return JSON objects.

Some operations run asynchronously. Video extraction, structured extraction, and site crawling return a job ID when processing continues in the background. The corresponding status tools accept that ID and report progress or results: transcript status, extract status, and crawl status each have a dedicated operation.

The default local transport is stdio. A hosted Streamable HTTP deployment is available at https://api.supadata.ai/mcp; it uses OAuth 2.1, and the /mcp route also accepts an x-api-key header. The repository includes a local development command for the HTTP worker.

Setup and configuration

For a local client, install and start the package with npx -y @supadata/mcp. Set SUPADATA_API_KEY to a key from the Supadata dashboard, either in the process environment or in the client’s MCP configuration. The sample configuration starts the command with npx, passes the package name as an argument, and supplies the key through env.

DEBUG=true enables verbose logging. RUN_STDIO=true forces the stdio server to start when its module is imported rather than executed directly. These settings are optional; the API key is required for the stdio server.

Integration instructions are provided for Claude, ChatGPT, Cursor, Windsurf, VS Code, and other clients in Supadata’s integration guide. Hosted OAuth sessions remain active while used and expire after 30 days without use. Codex users may need to run codex mcp login supadata after an expired login stops tool access.

Tools and capabilities

The Supadata MCP server includes these operations:

• supadata_transcript extracts a transcript from a supported video or file URL, with an optional language.
• supadata_check_transcript_status checks an asynchronous transcript job by ID.
• supadata_metadata returns platform, title, description, author, engagement, media, tags, and creation-date fields where available.
• supadata_extract uses a prompt, a JSON Schema, or both to derive structured data from a video; it returns a job ID for asynchronous processing.
• supadata_check_extract_status checks a structured-extraction job.
• supadata_scrape retrieves content from one URL with advanced options such as language.
• supadata_map lists indexed URLs discovered on a website.
• supadata_crawl starts extraction across multiple related pages and accepts a page limit.
• supadata_check_crawl_status checks a crawl job by ID.

Limitations and notes

The README identifies supported media platforms and file URLs, but does not define coverage for every URL type or describe rate limits, quotas, pricing, or extraction guarantees. Crawling and structured video extraction may require a separate status request because they are asynchronous. The hosted endpoint’s OAuth session behavior is relevant to long-lived client connections, particularly Codex sessions that stop working after expiration.