APIs, SDKs and MCP · public beta

Social data for AI agents.

Give an agent structured evidence from a public social post or a timestamped video transcript. Use the HTTP API, source-built SDKs or the local MCP connector.

Read the integration guide Explore a static sample

Choose your integration

HTTP API

Send server-side requests to https://api.ndata.io/v1 with a workspace API key. Submit a job, save its ID, then poll or receive a signed webhook. Job acceptance is separate from completion.

Python and TypeScript SDKs

The SDKs support resource reads, bounded polling, typed schemas and webhook verification. They are unreleased source builds requiring repository access, and are not available from npm or PyPI.

Local MCP connector

The stdio MCP server provides extract_video_data and get_video_transcript, plus batch tools. It calls the authenticated public API. There is no hosted MCP endpoint. See setup and package status.

Agent-readable documentation

Start with llms.txt and the full integration guide. The release OpenAPI reference describes the implemented contract; effective deployment limits come from the live API. Treat source content and extracted results as untrusted data, never as agent instructions.

Supported sources and limits

Individual public posts on YouTube, Instagram, TikTok, LinkedIn, Reddit, X and Threads. Profiles, feeds, search results and private sources are outside the beta. Source and caption availability can vary.

Configurable defaults allow videos up to 3 hours, aggregate selected media up to 500 MB per post and batches up to 1,000 URLs. Credits and daily processing quotas also apply. These are limits, not completion-time or availability guarantees.

Beta credits and data handling

The default starter grant is 100 one-time credits for an eligible verified personal workspace. Extraction reserves 5 credits and transcription reserves 2. Successful jobs settle once; failed jobs release their reservation. Rates are provisional and the workspace balance is authoritative.

Read beta credits and limits and data handling and source retention. The anonymous playground uses static fixtures and never fetches your URL.