Skip to main content
If Importly is connected to Claude Code or Cursor, the agent can also write the integration into your codebase. Your app uses the same API key and the same prices as the MCP server, through the REST API at https://api.importly.io.

Ask for it

Tell the agent what the feature should do. These prompts work as written:

Import from URL

Transcripts for every video

Link previews

The agent calls get_integration_guide with your stack (node, nextjs, python, or curl) and the use case. The guide costs nothing and returns code for each part of the integration.

What a good integration includes

Check the agent’s work against this list:
1

The key stays on the server

IMPORTLY_API_KEY is read from an environment variable in server code only. It’s never in client bundles or committed to the repo.
2

Jobs are started from your backend

A server route calls POST /import, /transcribe, or /metadata and stores the jobId with your record.
3

Results arrive by webhook

A webhook route receives the result and updates the record. Fields sit directly on data (data.mediaUrl, data.text), not under result. See Webhooks.
4

Polling covers failures

Failed jobs don’t send webhooks, so a fallback polls GET /<type>/status?id= every few seconds and records failed jobs with their message.
5

Errors are handled

402 means the balance is too low, and 429 means back off and retry. See Errors & rate limits.
mediaUrl links expire after 7 days. If your app keeps files, copy them to your own storage when the webhook arrives, or set store: true to deliver them to your bucket.

Without an agent

Follow the quick start, or use the API reference. Every page has a Copy page menu that gives you the Markdown version to paste into any AI tool.