Layer MCP

All of Layer, inside your agent

Connect any AI agent to Layer in under a minute and generate images, video, 3D, and audio with 301+ models — right where you already work.

Setup guide

OAuth, no API key. Works with Claude, Cursor, VS Code, Codex, and any MCP client.

Connect in under a minute

Pick your client and follow three steps. No API key, no local install for most clients.

1

Open Settings › Connectors in Claude and click Add custom connector.

Paste the Layer MCP server URL:


                      https://mcp.app.layer.ai/mcp

3

Click Add, then sign in with your Layer account when Claude prompts you.

Add the server from your terminal:


                      claude mcp add --transport http Layer https://mcp.app.layer.ai/mcp

2

Run /mcp, select Layer, and choose Authenticate.

1

Open Settings › Features › MCP and click Add New MCP Server.

Choose Streamable HTTP, name it Layer, and use this URL:


                      https://mcp.app.layer.ai/mcp

3

OAuth triggers automatically — sign in with your Layer account.

Add Layer to .vscode/mcp.json:


                      {
  "servers": {
    "Layer": {
      "type": "http",
      "url": "https://mcp.app.layer.ai/mcp"
    }
  }
}

2

OAuth triggers on first use — sign in with your Layer account.


                      codex mcp add layer -- npx -y mcp-remote@latest https://mcp.app.layer.ai/mcp

2

OAuth triggers on first use.

Add Layer to ~/.codeium/windsurf/mcp_config.json:


                      {
  "mcpServers": {
    "Layer": {
      "serverUrl": "https://mcp.app.layer.ai/mcp"
    }
  }
}

2

Verify it works by asking: "What can I do with Layer?"

Any client that can launch a command connects through the mcp-remote bridge:


                      {
  "mcpServers": {
    "Layer": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://mcp.app.layer.ai/mcp"]
    }
  }
}

2

Zed, Warp, Gemini CLI, Amp, OpenCode, OpenClaw, Hermes, and Manus have dedicated instructions in the setup docs.

Using Zed, Warp, Gemini CLI, Amp, OpenCode, OpenClaw, Hermes, or Manus? See all clients.

From conversation to finished asset

Every modality Layer supports is reachable as a tool call. Ask in plain language; your agent picks the model, prices the job, runs it, and hands back the files.

Images

Turn a sentence into finished art

Icons, key art, characters, environments, and UI — generated on transparent backgrounds or to your brief, in batches, without leaving the chat.

Video

Bring a still or an idea to life

Animate existing art or generate clips from a prompt, then assemble multi-shot timelines and render them — all through tool calls.

Reference image to 3D asset

Generate production-ready 3D models from a prompt or a reference image, ready to pull into Unity, Unreal, or Blender.

Audio

Sound effects and music on cue

Generate SFX, ambience, and music beds for prototypes and trailers in the same conversation as the visuals.

Prompts to start with

Paste one into a connected agent. It will choose the tools, estimate the cost, and show you the results.

Image

Icon variations

“Generate 4 variations of a crystal sword icon on a transparent background.”

3D from reference

“Create a 3D model of a treasure chest from this reference image.”

Workflow

Run a team workflow

“Run the product shot workflow with a photo of my item.”

Video

Animate a still

“Turn this key art into a 5-second looping video for our store page.”

Models

Model discovery

“Show me what image models are available for my workspace.”

Usage

Budget check

“How many Creative Units do I have left?”

21 tools, selected automatically

You never call these by hand. Your agent reads Layer's instructions first, then picks what it needs.

Models

Discover which generative models your workspace can use, and get recommendations for a task.

Forge

Run a single model directly with full control over its parameters.

Workflows

Execute the multi-step pipelines your team built in the Blueprint editor.

Video timeline

Assemble and render multi-shot video timelines from generated clips.

Files

Upload local files or public URLs to use as references and inputs.

Workspaces

List and inspect the Layer workspaces you belong to, including Creative Unit balance.

Instructions

Platform guidance the agent reads first, so it uses the rest of the toolset correctly.

Browse the full tool reference

Built for studio production, not demos

The same models, pipelines, budgets, and permissions your team already uses in Layer — reachable from an agent.

Your studio workflows, callable

The Blueprint pipelines your art directors already approved show up as tools. Your agent runs them with your inputs, so output stays on-style instead of drifting.

Costs estimated before they run

Every generation tool has a matching estimate tool. The agent can price a job in Creative Units and check your balance before spending anything.

One credit pool, no new seats

MCP usage draws on the same Creative Units as the app. Connecting an agent does not add per-seat cost for your team.

Workspace permissions still apply

The server authenticates as you over OAuth and respects your workspace roles. Admin tooling sits on a separate management server, out of the creative toolset.

MCP questions

What is MCP?

MCP (Model Context Protocol) is an open standard that lets AI assistants call external tools. Layer runs an MCP server, so any MCP-capable agent — Claude, Cursor, VS Code, Codex and others — can use Layer's generation tools directly in a conversation.

Do I need an API key?

No. Use https://mcp.app.layer.ai/mcp as the server URL and sign in with your Layer account when your agent asks you to connect. If your client does not support OAuth, you can authenticate with a Personal Access Token instead.

Which agents and editors are supported?

Claude (web, desktop, and Claude Code), Cursor, VS Code, Codex, Windsurf, Zed, Warp, Gemini CLI, Amp, OpenCode, OpenClaw, Hermes, and Manus all have step-by-step instructions in the setup docs. Any other MCP client can connect through the mcp-remote bridge.

What can I actually do through the server?

The creative server exposes 21 tools across models, forge (direct model runs), workflows, video timelines, files, and workspaces. Your agent picks the right ones automatically — see the full tool reference.

Which models are available?

The same catalogue as the Layer app — 301+ models across image, video, 3D, and audio, including day-zero integrations of new releases. Whatever your workspace can use in the app, your agent can use here. Browse the models.

How does billing work?

Generations run through MCP consume Creative Units from your workspace balance, exactly as they would in the app. There is no separate MCP charge and no extra seat cost. See pricing.

How fast is it?

Generation speed matches the underlying model. Long jobs run asynchronously — the agent starts the run, then polls for the result, so a slow video render does not block your conversation. Generation is rate limited to 60 requests per minute per user, plus 30 per minute for video.

Can I manage my workspace through MCP too?

Yes, through a separate Management MCP server that covers projects, groups, members, and usage. It is kept apart so admin tools do not crowd the creative toolset.

Is it available on my plan?

The MCP server is available to any Layer account. Enterprise workspaces keep their existing SSO, role-based access, and audit logging when agents connect. Talk to sales for a studio-wide rollout.

Still curious? Read why we built Layer MCP, or browse the full documentation.

Bring Layer into your agent

Connect in under a minute and start generating right where you already work.

Start for free Read the docs