{
  "markdown": "# Deepgram Skills\n\nSkills are folders of instructions and resources that AI coding tools load to help you build with Deepgram faster. Each skill teaches your tool how to use a specific part of the Deepgram platform — from API reference to finding the right starter app.\n\nFor more information, check out:\n- [Deepgram Documentation](https://developers.deepgram.com)\n- [Deepgram API Reference](https://developers.deepgram.com/reference/deepgram-api-overview)\n- [Ask Deepgram AI](https://developers.deepgram.com/ask-ai)\n\n# About This Repository\n\nThis repository contains skills for building with Deepgram's speech-to-text, text-to-speech, voice agent, and audio intelligence APIs. Skills are agent-agnostic — plain markdown that any AI coding tool can consume.\n\nDeepgram ships two actively maintained, industry-leading speech-to-text model families:\n\n- **Nova** (`/v1/listen`) — general-purpose transcription with a rich intelligence feature set (diarize, summarize, sentiment, topics, intents). Use for captions, subtitles, batch, and general live streaming.\n- **Flux STT** (`/v2/listen`, `model=flux-general-en`) — conversational STT with built-in turn detection. Use for voice agents and interactive assistants. See the `api` skill for the full Nova vs Flux STT decision guide.\n\nIt ships two text-to-speech families on separate endpoints — the voices do not overlap:\n\n- **Aura** (`/v1/speak`) — the broadest voice catalog (English, Spanish, German, Dutch, French, Italian, Japanese) plus compressed and containerized output. Use for one-shot synthesis and non-English voices.\n- **Flux TTS** (`/v2/speak`, `model=flux-{voice}-{language}`) — streaming-first, turn-based synthesis with barge-in and cross-turn voice consistency. Use for voice agents. `model` is required, and only `flux-*` voices are accepted. See the `api` skill for the full Aura vs Flux TTS decision guide.\n\nSome skills are hand-written, others are generated from Deepgram's [OpenAPI](https://dpgr.am/openapi.yml) and [AsyncAPI](https://dpgr.am/asyncapi.yml) specs.\n\n## Skills\n\n- [./skills](./skills): All Deepgram skills\n- [./template](./template): Skill template for creating new skills\n- [./scripts](./scripts): Scripts for fetching specs and generating skills\n\n# Skills\n\n| Skill | Description |\n|-------|-------------|\n| [speech-to-text](./skills/speech-to-text) | Start here for transcription: Nova on `/v1/listen` or Flux STT on `/v2/listen`, a first request, and where to go next |\n| [text-to-speech](./skills/text-to-speech) | Start here for synthesis: Aura on `/v1/speak` or Flux TTS on `/v2/speak`, a first request, and where to go next |\n| [voice-agent](./skills/voice-agent) | Start here for a voice agent: the Voice Agent API over one WebSocket, function calling, telephony wiring, and when to use an orchestrator instead |\n| [audio-intelligence](./skills/audio-intelligence) | Analyze audio, not just transcribe it: `summarize`, `sentiment`, `topics`, `intents`, and `detect_entities` as parameters on `/v1/listen` |\n| [text-intelligence](./skills/text-intelligence) | Analyze text you already have with the Read API: one `POST /v1/read` with `summarize`, `sentiment`, `topics`, and `intents` |\n| [browser-agent](./skills/browser-agent) | Run a voice agent in the browser with the Browser Agent SDK: `@deepgram/agents`, `@deepgram/react`, `@deepgram/ui`, and `@deepgram/agents-widget` |\n| [api](./skills/api) | Full API reference for all Deepgram REST and WebSocket APIs, generated from OpenAPI and AsyncAPI specs |\n| [docs](./skills/docs) | Find the right Deepgram documentation for any task |\n| [starters](./skills/starters) | Clone a ready-to-run demo app in your language and start building — 13 frameworks, 8 features |\n| [recipes](./skills/recipes) | Focused runnable recipes for one feature × one language — minimal working code (< 50 lines) |\n| [examples](./skills/examples) | Integration examples with third-party platforms (Twilio, LiveKit, LangChain, Vercel AI SDK, etc.) |\n| [cli](./skills/cli) | Drive Deepgram from the terminal with `deepctl`: install, auth, `dg listen`, `dg speak`, `dg init`, and where to reach for the API instead |\n| [setup-mcp](./skills/setup-mcp) | Set up the Deepgram MCP server for querying docs directly from your AI coding tool |\n| [self-hosted](./skills/self-hosted) | Run Deepgram on your own GPUs: licensing, container credentials, and per-target guidance for Docker/Podman, Kubernetes, and Amazon SageMaker |\n\n## SDK-Specific Skills\n\nEach Deepgram SDK repository publishes its own set of language-idiomatic skills under `.agents/skills/`:\n\n```bash\nnpx skills add deepgram/deepgram-python-sdk     # Python\nnpx skills add deepgram/deepgram-js-sdk         # JavaScript / TypeScript\nnpx skills add deepgram/deepgram-java-sdk       # Java\nnpx skills add deepgram/deepgram-go-sdk         # Go\nnpx skills add deepgram/deepgram-rust-sdk       # Rust\nnpx skills add deepgram/deepgram-dotnet-sdk     # C# / .NET\n```\n\nThe Swift and Kotlin SDK repositories are not public, so `npx skills add`\ncannot reach them; their skills will be listed here once the repositories open.\n\nFor browser work, use the `browser-agent` skill listed above. It covers the\nfour Browser Agent SDK packages published on npm:\n`@deepgram/agents`, `@deepgram/react`, `@deepgram/ui`, and\n`@deepgram/agents-widget`.\n\nEach SDK ships 7 product skills named `deepgram-{lang}-{product}`. The full set for the Python SDK:\n\n- `deepgram-python-speech-to-text`\n- `deepgram-python-text-to-speech`\n- `deepgram-python-text-intelligence`\n- `deepgram-python-audio-intelligence`\n- `deepgram-python-voice-agent`\n- `deepgram-python-conversational-stt`\n- `deepgram-python-management-api`\n\nThe `deepgram-{lang}-` prefix keeps names globally unique so installing skills from multiple SDKs never overwrites another SDK's skills.\n\nThis `deepgram/skills` repo covers product contracts (API reference, docs, starters, recipes, integrations, MCP). The SDK repos cover language-specific usage.\n\n# Install\n\n## Any AI coding tool\n\nWorks with Claude Code, OpenAI Codex, Cursor, Windsurf, GitHub Copilot, Gemini CLI, and [70+ others](https://github.com/vercel-labs/skills).\n\nInstall the skill for what you are building:\n\n```bash\nnpx skills add deepgram/skills --skill speech-to-text\nnpx skills add deepgram/skills --skill text-to-speech\nnpx skills add deepgram/skills --skill voice-agent\n```\n\nOr install every skill in this repository:\n\n```bash\nnpx skills add deepgram/skills\n```\n\nThe installer detects the agents on your machine. To target agents explicitly, pass `-a` with the agent name (`claude-code`, `codex`, `cursor`), and `--skill` to install one skill:\n\n```bash\nnpx skills add deepgram/skills -a codex -a cursor --skill api\n```\n\n## Claude Code (native plugin system)\n\nRegister this repository as a plugin marketplace:\n\n```\n/plugin marketplace add deepgram/skills\n```\n\nThen install the Deepgram plugin:\n\n```\n/plugin install deepgram@deepgram-agent-skills\n```\n\nThis gives you the following slash commands:\n\n- `/deepgram:speech-to-text`: Start here for transcription\n- `/deepgram:text-to-speech`: Start here for synthesis\n- `/deepgram:voice-agent`: Start here for a voice agent\n- `/deepgram:audio-intelligence`: Analyze audio on `/v1/listen`\n- `/deepgram:text-intelligence`: Analyze text with the Read API\n- `/deepgram:browser-agent`: Run a voice agent in the browser\n- `/deepgram:api`: Deepgram API reference\n- `/deepgram:docs`: Find the right documentation\n- `/deepgram:starters`: Clone a starter app\n- `/deepgram:recipes`: Focused runnable recipes for one feature × one language\n- `/deepgram:examples`: Integration examples with third-party platforms\n- `/deepgram:cli`: Drive Deepgram from the terminal with `deepctl`\n- `/deepgram:setup-mcp`: Set up the Deepgram MCP server\n- `/deepgram:self-hosted`: Run Deepgram on your own GPUs\n\nYou can also install SDK-specific skill plugins from the same marketplace:\n\n```\n/plugin install deepgram-js-sdk@deepgram-agent-skills\n/plugin install deepgram-python-sdk@deepgram-agent-skills\n/plugin install deepgram-java-sdk@deepgram-agent-skills\n/plugin install deepgram-go-sdk@deepgram-agent-skills\n/plugin install deepgram-rust-sdk@deepgram-agent-skills\n/plugin install deepgram-dotnet-sdk@deepgram-agent-skills\n```\n\nEach SDK plugin installs the 7 language-idiomatic skills from that SDK's repository.\n\n## OpenAI Codex\n\nCodex loads skills from `.agents/skills/` in your working directory and every directory up to the repository root. Install the Deepgram skills into a project:\n\n```bash\nnpx skills add deepgram/skills -a codex\n```\n\nThis writes `.agents/skills/<skill>/` for each skill you select. Codex also reads `~/.agents/skills/` for user-wide skills, so add `-g` to install there instead of into the project:\n\n```bash\nnpx skills add deepgram/skills -a codex -g\n```\n\nInside Codex, run `/skills` to browse and apply a skill, or type `$speech-to-text`, `$text-to-speech`, `$voice-agent`, `$audio-intelligence`, `$text-intelligence`, `$browser-agent`, `$api`, `$docs`, `$starters`, `$recipes`, `$examples`, `$cli`, `$setup-mcp`, or `$self-hosted` to invoke one by name. Codex also picks a skill on its own when your task matches its description. Codex reads a project's `AGENTS.md` before working. Use it to tell Codex when to use the Deepgram skills already installed in `.agents/skills/`.\n\nTo give Codex the live documentation as well, add the Deepgram docs MCP server.\n`api.dx.deepgram.com/kapa/mcp` rejects an unauthenticated request with HTTP 401\nand accepts a Deepgram API key as a bearer token, so pass the key through\n`--bearer-token-env-var`:\n\n```bash\ncodex mcp add deepgram-docs --url https://api.dx.deepgram.com/kapa/mcp \\\n  --bearer-token-env-var DEEPGRAM_API_KEY\n```\n\n`https://developers.deepgram.com/_mcp/server` needs no credential at all. The\n`setup-mcp` skill covers every path and the tools each one exposes.\n\nSee [Build skills](https://learn.chatgpt.com/docs/build-skills) and [Custom instructions with AGENTS.md](https://learn.chatgpt.com/docs/agent-configuration/agents-md) in the Codex documentation.\n\n## Cursor\n\nThe same installer targets Cursor, writing each skill to `.agents/skills/<skill>/` in the project:\n\n```bash\nnpx skills add deepgram/skills -a cursor\n```\n\n# Creating a Skill\n\nSkills are a folder with a `SKILL.md` file containing YAML frontmatter and instructions. Use the [template](./template/SKILL.md) as a starting point, and read [CONTRIBUTING.md](./CONTRIBUTING.md) for the folder layout, local testing, the review checklist, and the release process:\n\n```markdown\n---\nname: my-skill-name\ndescription: A clear description of what this skill does and when to use it.\n---\n\n# My Skill Name\n\n[Instructions that your AI tool will follow when this skill is active]\n```\n\nThe frontmatter requires two fields:\n- `name` — A unique identifier for your skill (lowercase, hyphens for spaces)\n- `description` — What the skill does and when to use it\n\n## Generating API Reference Skills\n\nThe API reference skills are generated from Deepgram's specs. To regenerate:\n\n```sh\n# Fetch the latest specs\nbun run scripts/fetch-specs.ts https://dpgr.am/openapi.yml https://dpgr.am/asyncapi.yml\n\n# Generate reference markdown\nbun install && bun run scripts/generate-skills.ts\n```\n",
  "bytes": 11194,
  "sha": "61e98688cb3fc4a9a12a65d6ea948490a961bbeec30785cf34cae09ef1d95383",
  "repo_slug": "deepgram/skills",
  "fonte": "repo",
  "truncated": false,
  "api": "https://api.agentalog.com/api/listings/skl_deepgram_skills_examples_7169bbaa/readme"
}