How to Add AI Music Generation to Gemini CLI with Suno MCP

If you already use Gemini CLI as a terminal assistant, the next useful step is not another chat window. It is giving that same terminal session a few real tools, so it can help you produce assets while you are still in the flow of building.

This guide shows how to add Suno MCP to Gemini CLI through Ace Data Cloud. The practical result is simple: from a Gemini CLI conversation, you can ask for a short song, background music, lyrics, a cover, or an extension of an existing track.

What you can do

The source document describes a remote MCP server for Suno. Once it is configured in Gemini CLI, the assistant can call Suno-related tools from natural language prompts instead of making you switch to a separate UI.

The documented MCP endpoint is:

https://suno.mcp.acedata.cloud/mcp

Authentication is passed with an HTTP header:

Authorization: Bearer yourToken

The tool list in the document includes:

Tool	What it is for
generate	Generate a song from a description
custom	Generate a song with custom lyrics and style
extend	Extend an existing song
cover	Cover a song in a new style
lyrics	Generate lyrics only
upload_to_persona	Register a Persona with vocal audio
query	Query historical generation results

That is enough for a practical builder workflow: write a tutorial, app demo, or game prototype, then create a matching audio draft without leaving the terminal.

How it works

Gemini CLI supports MCP servers. An MCP server is a tool provider that the model can call when it decides a task needs an external capability. In this case, the tool provider is the Suno MCP server hosted behind Ace Data Cloud.

The configuration has two parts:

1. the remote MCP URL
2. the authorization header that carries your Ace Data Cloud token

The one-command setup from the document is:

gemini mcp add suno \
  --transport http \
  https://suno.mcp.acedata.cloud/mcp \
  --header "Authorization: Bearer yourToken"

Use a real token locally, but keep examples and documentation in the yourToken form. Tokens should not be committed to a repository, pasted into public issues, or embedded in screenshots.

Alternative: edit the Gemini settings file

If you prefer declarative configuration, the document also shows a ~/.gemini/settings.json entry. This is useful when you want a repeatable setup that is easy to review.

{
  "mcpServers": {
    "suno": {
      "httpUrl": "https://suno.mcp.acedata.cloud/mcp",
      "headers": { "Authorization": "Bearer yourToken" }
    }
  }
}

After saving the file, restart your gemini session so the MCP server is loaded. I usually prefer starting with the CLI command for a quick test, then moving to the settings file once I know the endpoint and token are correct.

Try a few useful prompts

Once the server is mounted, the interaction can stay high level. You describe the musical goal, and Gemini CLI can choose the relevant Suno tool.

For a short social or product demo asset:

Help me write a 30-second upbeat Chinese pop song with the theme "Camping with friends on the weekend."

For background music under a screen recording:

Use Suno to create a lofi hip hop BGM for a programming tutorial video, without vocals.

For iterating on an existing track:

Extend this song (link xxx) from 30 seconds to 90 seconds, maintaining the original melody style.

These are intentionally specific. “Make some music” is hard to evaluate. “30-second upbeat Chinese pop song,” “lofi hip hop BGM,” and “extend from 30 seconds to 90 seconds” give the tool enough context to produce something you can review.

A practical workflow for builders

A small but useful workflow looks like this:

1. Write the script or demo flow in your project.
2. Ask Gemini CLI to summarize the mood and pacing of the content.
3. Turn that summary into a Suno prompt.
4. Generate one or two drafts.
5. If a draft is close, use extend or cover instead of starting over.
6. Use query when you need to check previous generation results.

The key is to treat audio generation like an iteration loop, not a one-shot command. The first output tells you what to clarify: vocals or no vocals, duration, genre, language, and whether the piece should sit under narration or stand on its own.

When I would use each tool

Use generate when you have a style description but no lyrics. This is a good fit for BGM, short loops, or mood boards.

Use custom when the words matter. If your demo video has a theme, tagline, or narrative arc, write or generate lyrics first, then pass the lyrics plus a style direction.

Use lyrics when you are still drafting the concept. It lets you separate writing from audio production.

Use extend when the first result is directionally right but too short. This is common for tutorial intros, product walkthroughs, and background tracks.

Use cover when the composition is useful but the style is not. Instead of discarding the idea, try a new style pass.

Use upload_to_persona only when you have appropriate rights to the vocal audio you are registering. That is an implementation detail, but it is also an important product boundary.

Keep the integration boring

For production or team usage, keep the setup boring and explicit. Store the token outside version control, document the MCP URL, and write down which prompts produce good results for your content style. If multiple people use the same workflow, standardize a few prompt templates so outputs are easier to compare.

This is where MCP feels natural: Gemini CLI stays the workspace, while Suno becomes one callable capability inside that workspace. You do not need a giant automation system to benefit from it; even a single terminal command plus a few disciplined prompts is enough to make the workflow useful.

You can read the original setup notes here: Gemini CLI with Suno MCP.