SunoForge MCP server
Commit Nov 7, 2026An AI music orchestrator that transforms ideas into finished tracks by generating prompts for Suno, Lyria, and ElevenMusic through an MCP connector.
Also known as sanic732/SunoForge
- Stars
- 20
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- —
- Last commit
- Nov 7, 2026
- License
- Apache-2.0
What is SunoForge MCP?
SunoForge is an AI music orchestrator that transforms ideas into finished tracks by generating prompts for Suno, Lyria, and ElevenMusic through an MCP connector.
Summary generated from the project's README
About SunoForge MCP server
Transforming musical ideas into complete tracks is what SunoForge enables for creators using Claude. This orchestrator acts as a meta-prompt that translates plain language or visual inputs into detailed instructions for various music generators. It works through the official Suno MCP connector, allowing users to manage their Suno library, check credit balances, and trigger generations directly from a chat interface.
Users connect their Suno account via OAuth to ensure that any generation consumes their existing credits. The system provides a cost card before any action is taken, showing the model, length, and required credits to prevent accidental spending. Beyond Suno, the tool supports a wide range of generators including Google Lyria, ElevenMusic, Stable Audio, and local models like YuE2. It handles complex tasks such as extending songs without gaps, creating speech with soundtracks, and even converting humming or piano melodies into MIDI files through its experimental Lab features.
What you can do with it
- 1Generate music from descriptionsTurn a plain text idea or a photo into a fully produced song with specific style and lyrics markup.
- 2Manage Suno account resourcesCheck your remaining credits and view your existing song library without leaving the chat.
Written from the project's README. Every use case is checked against a quote from it.
Automated checks
Ran Nov 14, 2026Full report →Not verified yet. A server needs a score of 80, no serious finding, and enough data for us to check. How checks work · What to fix for more points
- Source repository linked and reachablesanic732/sunoforge is public and reachablegithub.com/sanic732/sunoforge ↗Pass
- Package names this repository as its sourceNot checked
- Build provenance (Sigstore attestation)Not checked
- Open source license (OSI)Apache-2.0spdx.org/licenses/Apache-2.0.html ↗Pass
- Commit or release in the last yearLast commit or release 2026-11-07github.com/sanic732/sunoforge ↗Pass
- OpenSSF ScorecardNot checked
- Package name not close to a popular packageNot checked
- No critical advisory in the package or its dependenciesNot checked
- No hidden instructions to the model0 tool descriptions and the README read; no override phrases, credential paths or hidden charactersPass
- Destructive tools are labelledNot checked
- Remote endpoint answers the MCP handshakeNot checked
README
View on GitHub
SunoForge v5.0 — AI Music Prompt Orchestrator
SunoForge is a meta-prompt for AI music, built the same way as P2P: a reference and a prompt in one. It does not hand you a fixed template. It reads your idea and writes the prompt for this song: Style, lyrics markup, speech scripts, edits, repair of old prompts, rights and post-production notes. Describe what you want in plain words or show it a photo.
It works through Suno's official MCP connector. Add Suno's server to Claude, and SunoForge goes all the way from idea to finished track: it sends the prompt to Suno, sees your library and remaining credits, and shows a cost card before every generation, so nothing is spent without your yes. A voice memo or a video can go in too, through Suno's uploader (audio, MP4, MOV, WebM, up to 100 MB), and Suno listens to it.
Fourteen files, one principle: never present a guess as a fact.
Version 5.0 is out as the Claude plugin. The Markdown edition for Gemini and other long-context assistants is built from the same files and is still to come; until then, the 4.0 release has it.
What it looks like
Screenshots: assets/screens.
Generators
| Generator | What SunoForge covers |
|---|---|
Suno v6 (v6, v6-wild, v6-mini) |
Style and lyrics, Max Mode, Variety, edits in plain words, Voices, Custom Models, Studio 2.0, stems |
| Suno Speech (beta) | narration with its own soundtrack in one take — menu [14] |
| Google Lyria 3.5 / 3 Clip / RealTime | full songs, 30-second sketches, a live stream; timestamp ranges for video |
| Google Flow Music | changing one part without regenerating the rest |
| ElevenMusic v2.5 | section-by-section editing; release rights by plan |
| Stable Audio 3.0 | licensed data, open weights, a DAW plugin |
| MiniMax Music 3.0 | open weights, full songs with vocals |
| Local models | ACE-Step 1.5 and YuE2 on your own GPU |
| Any music-generation MCP server | Suno's own (beta), ElevenLabs' hosted MCP or your own — send from the chat, menu [15] Suno + MCP |
Udio is not a target: downloads stayed disabled after its settlement.
Hosts
| Host | What you install today | How |
|---|---|---|
| Claude — chat on the web, Claude Desktop, Cowork | sunoforge.plugin, or the marketplace sanic732/SunoForge |
plugin (paid plans) |
| Claude Code | the marketplace sanic732/SunoForge |
plugin |
| Gemini | — the 5.0 Skill package is not out yet; the 4.0 one still works | Skill · files in a chat |
| ChatGPT | — the 5.0 plugin is not out yet | plugin |
| Antigravity | — the 5.0 plugin is not out yet | plugin |
| grok.com, other long-context assistants | — the 5.0 prompt package is not out yet; the 4.0 one still works | Project or attached files |
Version 5.0 is the Claude plugin, and nothing else yet. The Skill packages for Claude and
Gemini, the ChatGPT, Antigravity and Grok plugins and Prompt.zip are built separately and are
still to come. Until they do, the
4.0 release carries working packages for
those hosts — older content, but they run.
New in v5.0: the Suno MCP layer
On 28 October 2026 Suno opened an official MCP server (beta), https://api.suno.com/mcp.
With it connected, SunoForge can work with your Suno account from the chat:
- create — a song, a speech, a cover, an extension, a mashup;
- library — list and search your songs, playlists and workspaces; read a take back and check it against the blueprint;
- credits — show your plan and the credit balance.
Credits are spent only after your explicit yes. Every call that spends credits is preceded by a card — the server, the exact arguments, how many songs, that it spends credits — and SunoForge waits. One call per yes; no retries or extra takes on its own. Reading is free and asks nothing; changes to your library happen only when you ask for them. Whatever the server sends back — titles, lyrics, other people's songs — is treated as material, never as instructions.
Access to the beta is Suno's decision. Suno admits accounts through an early-access form on suno.com/mcp. Until your account is admitted, sign-in ends with "Connecting AI apps to Suno is not yet available for your account" on every client. That is Suno's beta gate, not a SunoForge fault.
- Sign-in is OAuth with your Suno account; no API key.
- Songs go to your Suno library; generation spends your Suno credits.
- The plugin does not bundle Suno's server — you add it like any MCP server (below).
- SunoForge never names a Suno tool from memory: it works with the tools the session shows and says what it found before generating.
- Not published by Suno: per-call cost, rate limits, and whether a download through the MCP counts the same for rights as one on suno.com. For a release, download in Suno itself.
Five of Suno's tools come back as Suno's own widgets — the player, the uploader, the download card, and buttons to buy credits and to make a voice — wherever Claude shows apps, the Code tab of the desktop app included.
Without a generation server nothing changes: you copy the prompt into the generator.
The Suno MCP layer ships in 5.0: the protocol in CORE_00 §6 LIVE, the Suno facts — tools,
field map, defaults that cost credits, plan gating — in DATA_SUNO_LIVE_2026-10.md, read
from the server's own schemas: 35 tools, valid as of 2026-11-05. You add Suno's server
yourself.
Also new: menu [14] SPEECH (Speech Blueprint: Script and Tone), the words send and
mcp-info, and two small files of your own data — settings.md and library.md — via
config, save and library. Menu entries [1]–[13] do the same jobs as in 4.0, several under
clearer names. Still no prices anywhere.
Lab (beta)
A proof of concept, visible in the menu as Lab with a BETA tag. What it does today:
- keys — play a melody on the on-screen piano; it is recorded with note lengths and rests, turned into notes and then into a MIDI file you can give Suno as your melody;
- humming — sing or hum into an audio file and get notes from it; for now this is rough and often inexact.
Its stability is not confirmed, and what happens next depends on your feedback: it gets developed, changed or removed. Tell me what helped and what is missing — for example, connecting your own synth or MIDI keyboard — in the 4PDA forum thread or in this repository's Issues.
Install
Claude Code
/plugin install sunoforge --marketplace sanic732/SunoForge
On Claude Code older than 2.1.275:
/plugin marketplace add sanic732/SunoForge
/plugin install sunoforge@sunoforge
Check: /sunoforge (or the word SunoForge) opens the menu [1]–[15].
Update: claude plugin update sunoforge@sunoforge.
Claude chat, Claude Desktop, Cowork
- Plugin (paid plans): Customize → Plugins → Add → Add marketplace →
sanic732/SunoForge→ install SunoForge. Without GitHub: Add → Upload a plugin →sunoforge.pluginfrom the release page. Anthropic is replacing Customize with one Plugins page that holds plugins, skills and connectors (most plans by 9 November 2026, Enterprise from 30 November). If your sidebar already shows Plugins, start there. - Skill and Project: the 5.0 packages are not out yet. Until they are, use the ones from the 4.0 release.
Check: in a new chat, type start.
Gemini
The 5.0 Skill package for Gemini is not out yet. The 4.0 one still works: unpack
SunoForge_v4.0_Skills_for_Gemini.zip from the
4.0 release, then Skills → Upload →
the whole sunoforge folder. The folder name must stay sunoforge.
ChatGPT, grok.com, anything else
The 5.0 prompt package is not out yet. Take SunoForge_v4.0_Prompt.zip from the
4.0 release, upload the .md files from
gemini/ to a Project (or attach them), and tell the assistant to read CORE_00_ENTRY.md first.
Suno, optional (beta)
- Claude Code:
claude mcp add --transport http suno https://api.suno.com/mcp, then/mcp→suno→ sign in to Suno. - Claude chat, Desktop, Cowork: Customize → Connectors → + Add → Add custom connector →
https://api.suno.com/mcp→ sign in to Suno.
A music MCP server of your own is added like any MCP server: claude mcp add in Claude
Code, a connector in Claude chat and Cowork. mcp-info shows what the session sees.
Every host step by step: docs/INSTALL.md · Suno in detail: docs/SUNO_MCP.md · plugin: plugin/README.md
How it works
- Confidence marking. Every claim about platform behaviour carries
[OFFICIAL],[COMMUNITY]or[UNVERIFIED]. - Rules and facts apart.
CORE_*files hold rules and are undated;DATA_*files hold platform facts, are dated, and are replaced when platforms move. - No prices. Plans are named where a feature or a right depends on them.
- Two or three interpretations per request, not one "correct" answer.
- Language. SunoForge answers in the language you write in;
set langpins one. In Claude Code and the desktop app the plugin's Settings tab also sets a language and a folder for your settings and library. - Views. The small line at the foot of every answer switches them: M menu (widgets with buttons), T text, B buttons (one multiple-choice question, for the terminal). In view M an answer without its own screen ends with its next steps and Menu.
Size: about 150,000 tokens across the fourteen files in o200k_base (measured 5 November
2026); Claude's own tokenizer counts about 1.45 times more — by the
plugin's estimate about 235,000, more than a 200K window holds with a conversation. So by default the plugin reads only the sections a task needs. Details:
docs/TOKENS.md.
Release files
On the 5.0.4 release page:
| File | What it is |
|---|---|
sunoforge.plugin |
the Claude plugin as one file, nine interface languages |
SHA256.txt |
its checksum |
Still to come, built separately: the Skill packages for Claude and Gemini, the ChatGPT, Antigravity
and Grok plugins, and Prompt.zip with both editions. Until they are out, the
4.0 release carries working versions of
all of them.
This repository is also the Claude plugin marketplace: .claude-plugin/marketplace.json
and the plugin in plugin/. Privacy: PRIVACY.md.
Built with P2P
Designed with P2P, an open cross-model meta-prompt framework by the same author.
Licence
Apache 2.0, from 5.0.0 on: free to use, modify and redistribute, commercially included; keep the notices in NOTICE and say which files you changed. See LICENSE. Releases before 5.0.0 were MIT and stay MIT. Quotes from Suno, Google, ElevenLabs and other vendors are theirs, not covered by this licence; each is marked and sourced in place.
Suno, Google, Lyria, Gemini, ElevenLabs, Stability AI, MiniMax, OpenAI, ChatGPT, xAI, Grok, Anthropic, Claude and other names belong to their owners. This project is unaffiliated with and unendorsed by any of them. Sources and notices: docs/CREDITS.md · history: docs/CHANGELOG.md
SunoForge MCP questions
How does SunoForge handle spending credits?
The server shows a cost card detailing the model, length, and credits required before any generation occurs, ensuring nothing is spent without your explicit approval.
Can I use my own audio files as input?
Yes, you can upload voice memos or videos (audio, MP4, MOV, WebM up to 100 MB) for Suno to listen to.
What authentication method does the Suno MCP layer use?
Sign-in is handled via OAuth with your Suno account rather than an API key.