Process and Edit Audio with AI
Connect your AI assistant through Model Context Protocol (MCP) to use Neural Analog from your conversation. Describe the result you want: restore a recording, master a mix, separate stems, or turn audio into MIDI.
What can you do with the Neural Analog MCP?
Describe the result you want in your own words. Your assistant chooses the appropriate Neural Analog model or realtime effect, then gives you the result.
Clean up audio
Improve compressed, noisy, distorted, or AI-generated music from Suno, Udio, MP3 files, and more.
Master a mix
Balance loudness and tone, control peaks, and prepare a finished track for release.
Split or remove sounds
Separate vocals and instruments, isolate a sound, or remove crowd noise from a recording.
Turn audio into MIDI
Transcribe piano, drums, bass, guitar, vocals, or a full arrangement into a MIDI file.
Edit tracks and stems
Apply EQ, reverb, stereo width, panning, speed, or pitch to one track, many tracks, or selected stems.
Run complete audio workflows
Combine several steps—such as cleaning, stem splitting, effects, and mastering—in one request.
Import links and files
Start from supported music links, playlists, chat attachments, or files stored on your computer.
Reuse your Studio library
Refer to something you already imported or uploaded; your assistant can find its originals, restorations, masters, stems, and MIDI.
How to install it
Choose your assistant below. When its browser window opens, sign in to Neural Analog and approve the connection.
- Run this command in a terminal.
codex mcp add neuralanalog --url https://api.neuralanalog.com/mcp- Sign in to Neural Analog.
codex mcp login neuralanalogRestart Codex if it was already open. The Codex app, CLI, and IDE extension share this configuration on the same computer.
- Add Neural Analog for every project on your computer.
claude mcp add --transport http --scope user neuralanalog https://api.neuralanalog.com/mcp- Start Claude Code and enter the following command.
/mcpSelect neuralanalog, choose the sign-in action, and finish the authorization in your browser. You can also run claude mcp login neuralanalog from a terminal.
- Add this entry to
~/.config/opencode/opencode.json.
{
"$schema": "https://opencode.ai/config.json",
"mcp": {
"neuralanalog": {
"type": "remote",
"url": "https://api.neuralanalog.com/mcp",
"enabled": true
}
}
}- Open the browser sign-in flow.
opencode mcp auth neuralanalogYou can check the connection at any time with opencode mcp list.
- In ChatGPT on the web, open Settings → Security and login and enable Developer mode.
- Open Plugins and select the + button to add a developer-mode app.
- Name it Neural Analog, add a short description, and enter this MCP server URL.
https://api.neuralanalog.com/mcp- Choose OAuth, create the connection, and sign in to Neural Analog when prompted.
- In a new conversation, open the + menu, choose Developer mode, and select Neural Analog before making your request.
Developer mode requires an eligible account and workspace permissions. See the ChatGPT developer mode guide for current availability.
- Open Customize → Connectors in Claude.
- Select + → Add custom connector.
- Name it Neural Analog and paste the server URL.
https://api.neuralanalog.com/mcp- Select Add, then Connect and sign in to Neural Analog.
- In a conversation, use the + menu to enable Neural Analog before making your request.
On a Team or Enterprise plan, an organization owner must add the custom connector before members can connect it.
- Add a remote or Streamable HTTP MCP server.
- Paste the server URL below.
- Choose OAuth sign-in and approve Neural Analog.
https://api.neuralanalog.com/mcpThe client must support remote MCP servers and OAuth.
What can you ask?
Talk naturally and include the song link, attachment, local file path, or audio reference you want to process. Short and vague requests are fine.
“download this [Suno link]”
Your assistant imports the song into Neural Analog first, then downloads its original file.
“import these files [audio attachments]”
The files are added to your account for later processing. Successful imports remain available if another file fails.
“clean this up [Suno link]”
The song is imported, restored with the recommended settings for generated music, and returned as a clean audio file.
“make this sound better [local path to demo.mp3]”
The file is uploaded and restored. You do not need to choose a model or describe every defect.
“take the crowd out [local path to live-recording.mp4]”
The video's audio is separated and the music without the crowd is returned.
“master this and don't distort it [local path to mix.wav]”
The mix is mastered with controlled loudness and returned as a finished audio file.
“turn this into MIDI [local path to piano.mp3]”
The piano is transcribed with a piano-specific model and returned as a MIDI file.
“remove the vocals from this [public song link]”
The song is imported, split into stems, and returned without the vocal stem.
“process all the files in [folder]”
Every supported audio or video file in the folder is uploaded, processed with the requested or recommended preset, and returned in one ZIP.
“download all songs from [playlist]”
Your assistant imports the playlist within your plan's limits, then downloads the imported songs together in one ZIP.
“download all the midi files”
Your assistant uses the current song and saved Studio selection, then returns its selected MIDI parts together in a ZIP.
“download as midi”
Your assistant finds the relevant MIDI results from the conversation and Studio, returning a file or a ZIP as needed.
“clean up the last song I uploaded”
Your assistant finds the song in your private Studio library, restores it, and returns both the download and its Studio review page.
“split this into stems, boost the bass, then master the mix [local path to song.wav]”
The song is uploaded and split. A low-band EQ boost is applied only to the bass stem, then the current stems mix is mastered and returned as a finished file.
“clean this up, then master it [local path to rough-mix.mp3]”
A matching Neural Analog workflow runs every step in order and returns the actual final file, without asking you to manage each step.
“run my podcast workflow on everything in [folder]”
Your saved workflow is applied to every supported file in the folder, and the finished results are returned together.
Local files and URLs
You can import audio from local files or URLs.
Your account, minutes, and settings
The MCP uses your existing Neural Analog account, plan, processing minutes, and audio library. Connecting it does not start a job or consume processing minutes by itself.
Music-link and local-file imports follow your account's visibility rules. Downloadable attachments and direct audio file URLs use the private file-upload API. Review your account settings if needed.
To disconnect, remove Neural Analog from your assistant and revoke its access in your Neural Analog connected-app settings. This signs out that assistant without affecting your other Neural Analog sessions.