The Splice MCP Server lets AI agents search Splice’s sample library, create and edit multi-layer music arrangements, adjust BPM, swap sounds, and share or download audio samples.
Codex is OpenAI's coding agent for the terminal. It reads MCP servers from its config and signs in to remote ones with OAuth.
This guide connects Splice MCP to Codex through MewCP, so Codex can use Splice MCP's 6 tools, like create_stack, describe_a_sound and download_asset, without your Splice MCP credentials ever reaching the app.
Splice MCP signs in through managed OAuth. MewCP stores the token encrypted, refreshes it before it expires and injects it only at call time. Your agent never holds it, so it can never leak it.
More apps
Manage YouTube content including videos, channels, playlists and creator operations. Search content, organize media, publish updates and interact with YouTube resources programmatically.
Turn your AI agent into a creative studio The world's best image, video, audio and 3D models, inside the AI assistant you already use.
Give AI coding agents access to a curated library of real-world website designs, screenshots, design systems, and reusable UI components. Search visual references, find similar designs, analyze layouts, extract colors and typography, compare websites, and retrieve copy-ready JSX components for building better interfaces
Build a website or web app in minutes! Generate, design, write code, and create copy for your website.
Other clients
Splice MCP works the same way in every client below. Pick yours for its own guide.
Free to start.
Scope access to the exact Splice MCP tools and accounts you allow, work and personal side by side. Anything outside that scope is refused at call time, even when the model asks for it.
Each call is logged: what ran, on which account, and whether it worked. Failures come back structured, so the agent knows whether to retry instead of guessing.
Every app sits behind one secure MCP URL, and there is no server for you to deploy or keep running. Add an integration in the dashboard and it is live on the next call, with no config to change and no new credentials to wire up. Tools load only when a task needs them, so your agent stays fast and focused however many apps you connect.
Run this and approve access in the browser window.
Start a new chat and ask: "List the tools you can use from MewCP." You should see your connected apps.
Create a stack from an existing sample. This is the ONLY tool that can use a previously found or selected sample as a seed — if the user wants to build a stack around a specific sample, you MUST use this tool (not prompt_to_stack). ROUTING RULE: Any request like 'create a stack from this sample', 'build around this loop', 'use this as a seed', or referencing a specific asset UUID MUST use this tool. The seed sample MUST be a loop (not a one-shot). If the user asks to build a stack from a one-shot, do NOT silently search for loops. Instead, FIRST explain to the user that one-shots cannot be used as stack seeds because stacks require loopable samples to layer together, THEN offer to find similar loops that match the style or character of the one-shot they liked. Uses the loop as the starting layer and generates a full multi-layer stack of compatible samples around it.
Search the Splice catalog for sounds matching a natural language description. Returns sample metadata including BPM, key, tags, duration, and Splice links. Use this to BROWSE and EXPLORE individual sounds -- when the user wants to find, preview, or discover specific samples, loops, or one-shots. Do NOT call this tool alongside prompt_to_stack for the same request. If the user wants a complete multi-track beat or arrangement, use prompt_to_stack instead. Present ALL results to the user exactly as numbered; do not filter, omit, or reorder any results — let the user decide which sounds are relevant. IMPORTANT: One-shots CANNOT be used as seeds to create stacks — only loops can. If the user searched for one-shots and then asks to create a stack from those results, you MUST explicitly tell the user that one-shots cannot be used to create stacks and offer to search for similar loops instead. Do NOT silently pivot to loops.
Purchase and get a presigned download URL for a sound from the Splice catalog. The server returns a presigned URL — it does not write files. If the client has local filesystem access, it should fetch the URL and save the file to download_path; otherwise omit download_path and present the URL to the user. IMPORTANT: The first time an asset is downloaded it spends one of the user's credits. Subsequent downloads of the same asset are free. Always confirm with the user before calling this tool. The URL expires quickly — always display the file name alongside the download URL so the user can identify and access their file.
Generate a complete multi-track arrangement of compatible samples from a text description. Use this ONLY when the user describes a sound from scratch without referencing a specific existing sample. This tool does NOT accept a sample UUID as input -- it creates a stack entirely from a text prompt. ROUTING RULE: If the user says anything like 'create a stack from this sample', 'build a stack around this loop', or 'use this as a seed', you MUST use create_stack instead -- even if the user also provides a text description of the desired sound. Do NOT call this tool alongside describe_a_sound for the same request. If the user just wants to browse or find individual sounds, use describe_a_sound instead. Uses AI to interpret your prompt, select appropriate instrument layers (drums, bass, keys, etc.), and filters, and find matching harmonically coherent samples. BPM GUIDANCE: Always include a target BPM in the prompt AND set bpm_min/bpm_max to avoid extreme time-stretching. If the user specifies a genre but not a BPM, infer the range from the genre. Optionally generates a public shareable URL.
Generate a public shareable URL for an existing stack. Takes a stack UUID (from a previous create call) and returns a shareable link.
Modify an existing stack. Use this tool to add, remove, or swap sounds in a stack, rename it, or change its BPM. LAYER LIMIT: Stacks have a maximum of 8 layers. Adding layers beyond this limit will fail. If the stack already has 8 layers, remove existing layers before adding new ones. Required parameter: stack_uuid (the UUID, not a URL). To remove layers, pass their UUIDs via remove_layer_uuids. At least one modification must be provided. When adding or swapping layers, PREFER add_compatible_layers and swap_compatible_layers — these find harmonically compatible samples that fit the existing stack. Only use add_asset_uuids or swap_layers when the user has explicitly chosen a specific sample to use. DRUMS: When adding drums, ALWAYS include the 'grooves' tag in the tag_filter (e.g. '(hip hop || trap) && grooves', 'house && grooves') to get complete drum patterns (kick + snare + hats combined). Without the 'grooves' tag, the search returns isolated single-hit layers (just kicks, just snares, etc.) instead of full drum loops. Only omit 'grooves' if the user explicitly asks for an isolated drum element. Optional tag filter DSL to narrow sample selection. Prefer && (intersection) so all tags must match for precise results; fall back to || (union) only if intersection returns too few results. Valid tags per layer type: drums: grooves, house, hip hop, techno, tech house, hats, live sounds, trap, deep house, kicks, edm, pop, downtempo, snares, percussion, rnb; percussion: live sounds, house, techno, hip hop, tech house, cinematic, shakers, trap, downtempo, acoustic, deep house, edm, synth, pop; bass: synth, house, sub, techno, hip hop, tech house, edm, electric, trap, deep house, drum and bass, rnb, downtempo, dubstep; guitar: electric, live sounds, electric guitar, hip hop, acoustic, rnb, pop, rock, trap, soul, wet, funk, downtempo, lo-fi hip hop, dry; keys: piano, hip hop, electric piano, synth, rnb, soul, trap, live sounds, house, downtempo, pop, lo-fi hip hop, organ, deep house, jazz; strings: live sounds, hip hop, organic, acoustic, cinematic, synth, cello, folk, dry, clean, violin, trap, rnb, soul; vocals: female, pop, wet, edm, live sounds, dry, male, rnb, house, hip hop, pitched, cinematic, trap, soul, future bass; chords: synth, keys, hip hop, rnb, piano, guitar, live sounds, soul, trap, house, electric, pop, electric piano, downtempo, lo-fi hip hop; leads: synth, house, edm, wet, trap, techno, hip hop, pop, vocals, tech house, progressive house, rnb, guitar, deep house, trance; pads: synth, hip hop, house, downtempo, trap, techno, ambient, rnb, cinematic, pop, synthwave, deep house, edm, tech house, soul; brass-and-woodwinds: live sounds, hip hop, trap, organic, synth, cinematic, dry, acoustic, jazz, rnb, soul, clean; synth: house, techno, hip hop, trap, tech house, edm, deep house, downtempo, pop, rnb, progressive house, sub, wet, synthwave, future bass; fx: synth, techno, house, cinematic, ambient, edm, hip hop, vocals, percussion, downtempo, tech house, live sounds, trap, pop
