Connect Sonilo to your AI agent and generate licensed music, sound effects, and dubbing right from the conversation. Sign in from your agent; there is no API key to create.
The full Sonilo API as a CLI you can script, pipe, and drop into any CI job or Dockerfile. Install from npm or PyPI.
Eleven open-source skills that auto-trigger on what you say, covering scoring, SFX, dubbing, and ducking. Your agent picks the skill, the skill drives the API.
Open its MCP connector settings. The link below goes straight there.
Name it Sonilo and paste the URL below. Save it.
Connect, sign in with your Sonilo account, and ask your agent to score a video. No API key anywhere in the flow.
One command from npm or PyPI, with the same subcommands and the same output either way.
OAuth opens in your browser. The credential stays on your machine and every command picks it up, so there is nothing to paste and nothing to rotate.
Every endpoint is a subcommand. Async jobs poll themselves, so you just get a file back.
One command adds all eleven skills to any compatible assistant, or pick just the ones you need.
Skills are instructions; the MCP server is the tools they drive. Register it, then sign in once, and the credential covers the CLI too. No MCP host? Install the CLI and sign in: the skills fall back to it on their own.
Open the agent you added the skills to (Claude Code and Codex), signed in, and type this in its chat rather than a terminal. Skills auto-trigger on plain English, so describe the outcome with no tool names and no parameters.
One server hands your agent every Sonilo capability: scoring against footage, sound effects, combined soundtracks, ducking, and dubbing. One command registers it with your host, and it authenticates from the sign-in you already did, so there is no API key to create, paste or rotate.
claude mcp add sonilo -- uvx sonilo-mcpcodex mcp add sonilo -- uvx sonilo-mcpSign in once with sonilo login and both hosts share the credential. Prefer to hold a key? Add --env SONILO_API_KEY=sk-.... Prefer nothing local? Paste https://api.sonilo.com/mcp into your host's connector settings instead. Open docs →
Real output, generated by the Sonilo MCP server.
Ten generation capabilities plus three account tools. Each one is a single MCP tool call for your agent, a single command in your shell, or just a sentence once the skills are installed.
video_to_music“Score this clip so the music follows the cuts”
text_to_music“Make 30 seconds of warm lo-fi piano with rain underneath”
video_to_video_music“Add a soundtrack to this clip and give me the video back”
video_to_sfx“Add sound effects for whatever happens in this clip”
text_to_sfx“I need a glass bottle shattering on concrete”
video_to_video_sfx“Mix effects into this clip and return the video”
video_to_sound“Give this clip music and effects in one balanced mix”
video_to_video_sound“Full soundtrack on this clip, muxed back into the video”
audio_ducking“Duck this music under the voiceover”
dubbing“Dub this into Japanese and Spanish, keep the music bed”
get_generation_taskPoll async jobs
get_usageSpend & quota
get_account_servicesWhat your key can do
One request can move through several tools. Your agent picks the chain from what you asked for; from the CLI you run the same steps in order.
“Score this cut and keep the dialogue on top of the music.”
cut.mp4video_to_musicaudio_duckingscored.m4a“Add a soundtrack, then give me Japanese and Spanish versions.”
spot.mp4video_to_video_musicdubbingspot-ja.mp4 · spot-es.mp4“Add foley to every clip in ./renders.”
renders/*.mp4video_to_sfx×12renders/*.m4a