Skip to content

audiopodai/audiopod

v1.0.0MIT

Audio AI for agents: text-to-speech, voice cloning, music generation, stem separation, speaker separation, transcription, denoise and media conversion over AudioPod's hosted MCP server, plus 14 task skills.

AudioPod agent plugin

One installable package that gives any plugin-capable coding agent AudioPod's audio AI β€” ten MCP tools over a single hosted endpoint, plus fourteen task skills that describe the full multi-step flows (narrate an audiobook, build a podcast, cut a karaoke track).

No install, no local models, no GPU.

MCP endpointhttps://mcp.audiopod.ai
TransportStreamable HTTP (MCP 2025-06-18)
AuthX-API-Key: ap_* (or OAuth2 / JWT) β€” get a free key
Docshttps://docs.audiopod.ai/sdks/mcp
Pricinghttps://www.audiopod.ai/developers/pricing
Registryai.audiopod/audiopod on the official MCP Registry
Server repoAudiopodAI/audiopod-mcp β€” the standalone MCP listing (server.json + server card)

Your first API key comes with $1 of free API credit.

Set your key first

Every client below reads the key from the AUDIOPOD_API_KEY environment variable β€” except Gemini CLI, which prompts for it and stores it in your system keychain. No key is ever written into this repo.

export AUDIOPOD_API_KEY=ap_your_key_here

Install

Claude Code

claude plugin marketplace add AudiopodAI/audiopod-plugins
claude plugin install audiopod@audiopod

Then /audiopod in a session to see what is available and check the connection. The owner/repo shorthand falls back to HTTPS when no GitHub SSH key is configured; claude plugin marketplace add https://github.com/AudiopodAI/audiopod-plugins is the explicit equivalent.

Gemini CLI

gemini extensions install https://github.com/AudiopodAI/audiopod-plugins

Gemini reads gemini-extension.json, prompts for the AudioPod API key on install and stores it as a sensitive setting.

Codex

codex plugin marketplace add AudiopodAI/audiopod-plugins
codex plugin add audiopod@audiopod

Codex reads the same marketplace manifest this repo ships for Claude Code, then loads plugin.json, mcp.json and skills/ from the plugin root. /plugins inside Codex CLI browses configured and remote marketplaces instead.

Cursor

Customize β†’ Plugins β†’ search for AudioPod β†’ Install, and pick project or user scope. To test before it is listed, clone the repo into ~/.cursor/plugins/local:

git clone https://github.com/AudiopodAI/audiopod-plugins.git ~/.cursor/plugins/local/audiopod

Kiro

Powers panel β†’ Add Custom Power β†’ Import power from GitHub, enter https://github.com/AudiopodAI/audiopod-plugins, then Install.

Any other MCP client

Skip the plugin layer and point your client straight at the server:

{
  "mcpServers": {
    "audiopod": {
      "type": "streamable-http",
      "url": "https://mcp.audiopod.ai",
      "headers": { "X-API-Key": "ap_YOUR_KEY" }
    }
  }
}

Tools

ToolDoes
text_to_speechSpeech in 200+ languages, 500+ voices and custom clones
clone_voiceClone a voice from a 5–30s reference clip
change_voiceConvert a recording to a different target voice
generate_musicSongs, instrumentals, rap, or vocal stems from a text prompt
separate_stemsSplit a track into stems (vocals, drums, bass, …); two-stem mode = karaoke
separate_speakersIsolate each speaker into a separate track
transcribe_audioTranscribe with word-level timestamps and speaker diarization
denoise_audioRemove background noise while preserving voice character
convert_mediaConvert audio/video formats (mp3/wav/flac/ogg/m4a, mp4/mov)
check_job_statusPoll the status/result of a long-running job (free)

Long-running tools return a job_id immediately β€” poll check_job_status until COMPLETED. Polling is free.

Skills

Fourteen skills ship in skills/, each a SKILL.md the agent loads on demand. They cover the tasks that need more than one call, and document the REST endpoints for the ones that are not exposed as MCP tools.

SkillFor
text-to-speechVoiceover and narration, with inline directing and word timestamps
design-voiceBuild a brand-new voice from a text description
clone-voiceClone a voice from a reference recording
change-voiceRe-voice an existing recording
narrate-audiobookManuscript β†’ ACX-spec audiobook with chapters and retail sample
create-podcastScript or source material β†’ multi-voice podcast episode
read-articleArticle or URL β†’ listenable audio
generate-musicSongs, instrumentals and stems from a prompt
separate-stemsStems and karaoke tracks
separate-speakersOne track per speaker
transcribe-audioTranscripts with timestamps and diarization
translate-audioTranslate and re-voice into another language
denoise-audioClean up noisy recordings
convert-mediaFormat and container conversion

What is in this repo

PathRead by
plugin.json, mcp.jsonAgent Plugins layout β€” Codex, Cursor, Kiro
.claude-plugin/plugin.json, .claude-plugin/marketplace.jsonClaude Code
.cursor-plugin/plugin.jsonCursor
gemini-extension.json, GEMINI.mdGemini CLI
skills/<name>/SKILL.mdall of the above
commands/audiopod.mdClaude Code β€” the /audiopod slash command

The manifests are deliberately duplicated rather than merged: the formats agree on skills/ and on the shape of an MCP server entry, but disagree on where the manifest lives and on the key that carries the endpoint URL (url vs httpUrl).

Building a startup on this?

Apply to AudioPod for Startups β€” free Pro for 3 months plus developer API credits for eligible early-stage teams.

License

MIT β€” see LICENSE. (Covers this plugin package; the AudioPod service itself is governed by the AudioPod Terms.)