REPO PACKAGE
deepgram deepgram-js-sdk
COMMUNITY7skills
0ready to use
LABSCO SUMMARY
The seven split Deepgram's API surface by endpoint: speech-to-text covers plain transcription on /v1/listen (REST file upload or WebSocket streaming); conversational-stt covers the newer, turn-aware Flux models on /v2/listen; text-to-speech covers one-shot and streaming synthesis on /v1/speak; audio-intelligence covers overlays on the same listen endpoint — summarize, topics, sentiment, diarize, redact, detect-language; a seventh, text-intelligence, runs those same analytics on text you've already transcribed instead of raw audio; voice-agent covers the full-duplex STT+LLM+TTS runtime at agent.deepgram.com; and management-api covers project, key, member, and billing administration. Each skill's own description names its neighbors and says when to use them instead — STT versus conversational STT, TTS versus voice agent — so an agent reaching for the wrong endpoint is the failure mode this package is built to prevent.
This is for developers already writing JavaScript or TypeScript against @deepgram/sdk who want an agent that calls the SDK's actual current methods — client.listen.v1.media.transcribeFile, client.agent.v1.createConnection, and so on — instead of guessing at a shape that changed between the v4 and v5 migrations the README documents. It has nothing to offer anyone not already using this SDK.
READ THE FULL ANALYSIS
What it costs to start. All seven skills assume a live Deepgram account: the SDK requires DEEPGRAM_API_KEY or a minted access token before any call any of these skills teach will return real data. There is no free or account-less skill in this package — every one of the seven is classified as needing that account and key.
Narrow by design. This only covers the JavaScript/TypeScript SDK — it says nothing about Deepgram's Python, Go, or other language clients, and nothing about using Deepgram's REST API directly without this SDK.
ALSO IN THIS PACKAGE
PROMPT TEMPLATES
1 itemPROMPT TEMPLATE
pr interface review
MCP SERVERS
1 itemMCP SERVER
deepgram-styles
stdio · npx -y -p @deepgram/styles mcpWHAT'S INSIDE
7 showing · 7 total7 of 7
SETUP
Showing all 7 skills
deepgram-js-audio-intelligence
Switching on a few extra options turns an ordinary Deepgram transcription into an analysis as well — a summary, the topics raised, the mood of each passage, and a label on every line saying which speaker said it.
deepgram-js-conversational-stt
In a spoken conversation the hard part is knowing when someone has actually finished talking — this Deepgram service writes down live audio and marks the end of each turn as it happens.
deepgram-js-management-api
You manage your Deepgram account from code: projects, API keys, team members, and usage and billing details, all without opening the dashboard.
deepgram-js-speech-to-text
The plainest of Deepgram's audio jobs: speech goes in, written words come out — from a finished recording handed over in one go, or from audio still streaming in.
deepgram-js-text-intelligence
Hand Deepgram writing you already have — a transcript, an email, a chat log — and it reports back what the text is about, the mood behind it, and a short summary.
deepgram-js-text-to-speech
You turn written text into spoken audio, either as a finished audio file or as a low-latency stream while the text is still being generated.
deepgram-js-voice-agent
Deepgram runs the whole back-and-forth of a spoken assistant: it hears the person talking, works out a reply with a language model, says it out loud, and can call your own code partway through.
No skill in this package matches that.
HOW TO GET IT
WHOLE PACKAGEInstall the whole package (7 skills)
npx skills add deepgram/deepgram-js-sdkSINGLE SKILLOr install a single skill
npx skills add deepgram/deepgram-js-sdk --skill <name> --full-depthPick the skill name from the Skills tab — each entry there installs independently.