# AudioPod AI — Developers

> Build on AudioPod over a remote MCP server, a REST API, or the Python and Node SDKs. Every capability below is a real, live route.

Canonical URL: https://audiopod.ai/developers
Last updated: 2026-05-21

Page: https://audiopod.ai/developers

<!-- GENERATED by pnpm gen:llms — edit src/lib/agents/llmsTxtAgentSection.ts, not this file. -->

## For Agents & Developers

AudioPod is **agent-native**: every audio capability is reachable over a remote MCP server, a REST API, or an SDK. There is nothing to install for MCP — it is a hosted URL.

### Remote MCP server

- **Endpoint**: https://mcp.audiopod.ai (Streamable HTTP, MCP protocol 2025-06-18)
- **Core endpoint**: https://mcp.audiopod.ai/core — only tools for audio you already have (`transcribe_audio`, `separate_stems`, `denoise_audio`, `convert_media`, `separate_speakers`, `audio_to_midi`, `check_job_status`, `get_account`, `get_profile`); generates no new audio
- **Auth**: an AudioPod API key, as `X-API-Key: ap_…` or `Authorization: Bearer ap_…`, or OAuth sign-in with an AudioPod account
- **Get a key**: https://audiopod.ai/dashboard/account/api-keys — new developer accounts start with $1 of free API credit
- **Works with**: any MCP client that can add a remote Streamable-HTTP server — desktop assistants, agent-connector directories, and MCP-aware editors and CLIs

Tools the server exposes, exactly as `tools/list` returns them:

- `generate_music` — AudioMusic is AudioPod's music generation engine: original songs, beats and instrumentals from a description, with synced lyrics. AudioPod writes the lyrics unless the user supplies them.
- `compose_music` — Write an editable score (melody, chords, sections) for a new song without recording it yet — for when the user wants to shape the tune first.
- `revise_music_score` — Change a composed score in plain words ("brighter chorus", "down a key"); everything you did not ask to change is kept. Available when enabled for the account.
- `record_music_score` — Record a score as a finished track, optionally with a new style or new words for the same tune.
- `get_music_score` — Read the editable score of a track or composed score you own: key, meter, tempo, sections, length and the notation. Read-only.
- `edit_music_score` — Deprecated alias of record_music_score: record a track from a complete replacement score.
- `text_to_speech` — AudioSonic is AudioPod's text-to-speech engine: natural, studio-grade narration with inline emotion directing, timed pauses, IPA pronunciation control, and optional word-level timestamps.
- `transcribe_audio` — AudioTranscribe is AudioPod's speech-to-text engine: transcription with speaker diarization, word-level timestamps, and standard or premium accuracy, plus real-time streaming.
- `separate_stems` — AudioStems is AudioPod's stem separation engine: split a song into vocals, drums, bass, guitar, piano and more, or isolate any of a 45-instrument catalog.
- `clone_voice` — Voice cloning runs on the AudioSonic engine: create a custom voice from a reference clip and reuse it for narration or text to speech. Only the user's own voice, or a voice whose speaker has given explicit permission; requests to clone a public figure or anyone else without consent are declined, and the user must confirm consent before cloning.
- `denoise_audio` — Remove background noise, hiss, hum, and room tone from a recording while preserving the voice character.
- `convert_media` — Convert audio and video between all major formats (MP3, WAV, FLAC, OGG, M4A, AAC, and common video containers).
- `separate_speakers` — AudioDiarize is AudioPod's speaker separation engine: it diarizes a multi-person recording and splits it into an individual audio track per speaker.
- `change_voice` — Convert a recording into a different target voice, preserving the words and timing of the original performance.
- `audio_to_midi` — Convert a recording of a melody, instrument or song into MIDI files (per-instrument MIDI plus one multitrack MIDI), for notation or DAW work.
- `check_job_status` — Poll a long-running job by the job_id an earlier tool call returned, and read its status and finished output URLs.
- `get_account` — Read the signed-in account's plan and credits left (monthly allowance and pay-as-you-go), with links to manage the account.
- `get_profile` — Identify the connected account with a stable, opaque account id.

Long-running tools return a `job_id` immediately. Poll `check_job_status` with that `job_id` until the status is `COMPLETED`, then read the output URLs from the result.

### REST API

- **Base URL**: https://api.audiopod.ai/api/v1
- **OpenAPI spec**: https://api.audiopod.ai/api/v1/openapi.json (mirrored at https://audiopod.ai/openapi.json)
- **Auth**: `X-API-Key: ap_…` or `Authorization: Bearer ap_…`
- **Docs**: https://docs.audiopod.ai
- **Versioning**: the version is in the path (`/api/v1`). Breaking changes ship as a new path version; the retiring version carries `Deprecation` and `Sunset` headers.

Primary route per capability:

| Capability | Primary route |
|---|---|
| Text to Speech · powered by AudioSonic | `POST /api/v1/voice/voices/{voice_id}/generate` |
| Voice Cloning · powered by AudioSonic | `POST /api/v1/voice/voice-clone` |
| Voice Design · powered by AudioSonic | `POST /api/v1/voice/design` |
| Music Generation · powered by AudioMusic | `POST /api/v1/music/text2music` |
| Transcription · powered by AudioTranscribe | `POST /api/v1/transcription/transcribe` |
| Speaker Separation · powered by AudioDiarize | `POST /api/v1/speaker/diarize` |
| Stem Separation · powered by AudioStems | `POST /api/v1/stem-extraction/extract` |
| Audiobook Narration | `POST /api/v1/audiobook/projects` |
| Podcast Creation | `POST /api/v1/podcast/projects` |
| Audio Reader | `POST /api/v1/reader/generate` |
| Noise Reduction | `POST /api/v1/denoiser/denoise` |
| Media Conversion | `POST /api/v1/media-converter/convert` |
| Audio Translation | `POST /api/v1/audio/translations` |

### API metering

Calls made with an API key (REST, the Python or Node SDK, the CLI, or MCP with an API key) are priced per call in USD and paid from a separate prepaid API wallet first; when the wallet cannot cover a job, your account's plan and pay-as-you-go credits pay that USD price at $1 = 7,500 credits. A call is refused (HTTP 402) only when both are empty. Top up the wallet at https://audiopod.ai/dashboard/account/api-keys.

MCP tool calls made after signing in with OAuth, and jobs run in the dashboard, use your account's AudioPod credits (plan and pay-as-you-go) at the normal credit rates.

- **Speech to text** — $0.01 per minute of audio
- **Text to speech** — $0.05 per 1,000 characters of input text
- **Audiobook narration** — $0.09 per 1,000 characters of manuscript text

The complete, live rate table is `GET https://api.audiopod.ai/api/v1/api-wallet/pricing`. Read it there rather than caching these numbers.

### SDKs and CLI

- **Python**: `pip install audiopod` — https://pypi.org/project/audiopod/
- **Node**: `npm install audiopod` — https://www.npmjs.com/package/audiopod
- **CLI**: bundled with each SDK — `audiopod music`, `audiopod tts`, `audiopod transcribe`, `audiopod stems`, `audiopod clone`, `audiopod jobs`

### Discovery documents

- `/.well-known/mcp.json` — MCP server card — transport, auth, and the tool list
- `/.well-known/mcp/server-card.json` — the same body at the original path
- `/.well-known/agent-card.json` — A2A agent card
- `/.well-known/agent-skills/index.json` — agent-skills index
- `/.well-known/api-catalog` — RFC 9727 linkset — spec, docs, status, auth, and the per-directory connector manifests
- `/.well-known/oauth-protected-resource` — RFC 9728 protected-resource metadata
- `/.well-known/openapi.json` — OpenAPI spec mirror
- `/skill.md` — one-shot onboarding brief an agent can fetch and follow

The developer landing page is https://audiopod.ai/for-agents, and a signed-in playground that runs one real API call against your own key is at https://audiopod.ai/dashboard/account/api-keys/playground.
