Atlas Cloud MCP Server

English | 中文 | 日本語 | 한국어 | Español | Français

Use Atlas Cloud's 300+ image / video / LLM models in Claude Code, Codex, Gemini CLI, Cursor, Cline and more. Generate images, videos & chat via standard MCP tools.

→ Get your free Atlas Cloud API key · 300+ models · OpenAI-compatible

---

Supported Models

• 🎬 Video (218) — MiniMax H3 Max · MiniMax H3 Fast · Gemini Omni 1.1 Flash · MiniMax H3 · Wan-3.0-Prime · Wan-3.0

• 🎨 Image (138) — GPT Image 2.5 Sunburst · GPT Image 2.5 Flare · Seedream v4.7 · MAI-Image-2.6-Flash

• 🧊 3D (14) — Seed3D 2.0 · Tripo H3.1 · Hunyuan 3D Rapid · Hunyuan 3D Pro

• 💬 LLM (71) — DeepSeek V4.1 Flash · GLM 5.3 Flash · DeepSeek V4 Pro 0813 · Grok 4.6

• 🔊 Audio (TTS · Music · ASR) (13) — Seed Audio 1.0 · MiniMax Speech 2.6 Turbo · MiniMax Speech 2.6 HD · xAI TTS v1

• 📚 Explore more — all 486 live models »

🎬 Newest video models — Seedance 2.5 · Kling 4.0 · Wan 3.0 · Kling Video O3.

Contents

• What You Can Do
• Quick Start
• Available Tools
• Usage Examples
• Development
• More Atlas Cloud Tools
• License

What You Can Do

Ask your AI assistant in plain language — it discovers the right model, builds the parameters, and submits the job:

• 🎨 "Make a hero image for this blog post" — text-to-image across Nano Banana Pro, GPT Image 2, Flux 2, Seedream, Imagen…
• 🎬 "Turn this product photo into a 5-second ad" — image-to-video with Seedance 2.5, Kling 3, Kling Video O3, Veo 3.1…
• 🧊 "Make a 3D model from this photo" — image-to-3D / text-to-3D with Hunyuan 3D (GLB/OBJ/USDZ output)
• 🔊 "Read this script aloud" — text-to-speech with Seed Audio, ElevenLabs, xAI TTS
• 🎵 "Write a theme song for my app" — music generation with Suno, MiniMax Music
• 📝 "Transcribe this meeting recording" — speech-to-text with Seed ASR, xAI STT
• 🎞️ "Storyboard this script into 6 shots" — chain LLM → image → video inside one conversation
• ✏️ "Edit this image — add a hat" — upload a local file, then run an image-editing model
• 💸 "How much credit is left, and what did I spend this month?" — check balance, usage, and cost breakdowns
• 💬 "What is in this screenshot?" — LLM chat with Claude, GPT, DeepSeek, Qwen, Gemini, GLM…, including image/video input on models that accept it
• 🕓 "What did I generate yesterday?" — browse generation history and pull back the output URLs

Under the hood: model discovery, dynamic per-model parameter schemas (validated before every request so invalid params fail fast without spending credits, plus a dry_run flag on every generation tool that shows the exact request body without submitting it), media upload, one-step quick-generate, generation history, account balance & usage, and documentation search — all exposed as standard MCP tools (see Available Tools).

LLM calls are protocol-aware: each model is called through the contract it actually declares — OpenAI chat completions, OpenAI Responses, Anthropic Messages, or native Gemini generateContent — so Gemini-native models work without any special handling on your side.

Quick Start

Hosted server — no API key, nothing to install

Point your client at our hosted MCP server and sign in once in the browser. Nothing to install, no key to create or paste, and generation is billed to the Atlas Cloud account you sign in with.

Endpoint: https://mcp.atlascloud.ai/mcp

Claude Code

claude mcp add --transport http --scope user atlas-cloud https://mcp.atlascloud.ai/mcp

Drop --scope user to add it to the current project only. Adding the server does not sign you in — run /mcp inside Claude Code afterwards and follow the browser prompt.

Or add it to .mcp.json directly:

{
  "mcpServers": {
    "atlas-cloud": {
      "type": "http",
      "url": "https://mcp.atlascloud.ai/mcp"
    }
  }
}

type is required here — without it Claude Code reads the entry as a stdio server and skips it.

ChatGPT & Codex

Install the plugin, which bundles this server together with our skills:

codex plugin marketplace add AtlasCloudAI/atlas-cloud-plugin
codex plugin add atlas-cloud@atlas-cloud
codex mcp login atlas-cloud

In the ChatGPT desktop app: Settings → Plugins → Add → Add plugin marketplace, source AtlasCloudAI/atlas-cloud-plugin, then open Atlas Cloud and install it.

VS Code

In .vscode/mcp.json:

{
  "servers": {
    "atlas-cloud": {
      "type": "http",
      "url": "https://mcp.atlascloud.ai/mcp"
    }
  }
}

Other MCP clients

Most take the endpoint under an mcpServers key. The exact field names differ between clients — some want the transport spelled out, others infer it from the URL — so check your client's own MCP documentation.

Authentication is OAuth 2.1: the client registers itself, opens your browser once, and stays authorized after that. There is no API key anywhere in this path.

---

Prefer to run the server on your own machine, with your own API key? The rest of this section covers that.

Prerequisites

• Node.js >= 18
• Atlas Cloud API Key — Get one free at atlascloud.ai

See .env.example for the environment variable to set.

CLI agents (one-line install)

The fastest path — these AI coding agents add the server with a single command:

# Claude Code
claude mcp add atlascloud -- npx -y atlascloud-mcp

# OpenAI Codex CLI
codex mcp add atlascloud -- npx -y atlascloud-mcp

# Gemini CLI
gemini mcp add atlascloud -- npx -y atlascloud-mcp

# Goose CLI
goose mcp add atlascloud -- npx -y atlascloud-mcp

Set the ATLASCLOUD_API_KEY environment variable in your shell first.

IDEs, editors & extensions (JSON config)

Add this to your client's MCP configuration — works with every MCP-compatible client:

{
  "mcpServers": {
    "atlascloud": {
      "command": "npx",
      "args": ["-y", "atlascloud-mcp"],
      "env": {
        "ATLASCLOUD_API_KEY": "your-api-key-here"
      }
    }
  }
}

Client	Where to add it
Cursor	Settings → MCP → Add Server
Cline	MCP Marketplace → Add Server
Continue	config.yaml → MCP
Windsurf	Settings → MCP → Add Server
VS Code (Copilot)	.vscode/mcp.json or Settings → MCP
Trae	Settings → MCP → Add Server
JetBrains IDEs	Settings → Tools → AI Assistant → MCP
ChatGPT Desktop	Settings → MCP
Amazon Q Developer	MCP Configuration
Roo Code	Settings → MCP → Add Server

Prefer Skills?

If you'd rather use Skills than MCP, we also ship an Atlas Cloud Skills package for Claude Code and other skill-compatible agents.

Available Tools

Tool	Description
atlas_search_docs	Search Atlas Cloud documentation and models by keyword
atlas_list_models	List available models, filtered by type (Text/Image/Video/Audio), sub-kind (3d/tts/stt/music/lyrics) or keyword
atlas_get_model_info	Get detailed model info including API schema, parameters, and usage examples
atlas_generate_image	Generate images and 3D models (image-to-3D / text-to-3D) with any supported Image model
atlas_generate_video	Generate videos with any supported video model
atlas_generate_audio	Generate audio — speech (TTS) and music/songs (Suno, MiniMax Music) — with any supported audio model
atlas_transcribe_audio	Transcribe speech to text (ASR) — meetings, interviews, voice notes
atlas_quick_generate	One-step image/video/audio generation — auto-finds model by keyword, builds params, and submits
atlas_upload_media	Upload a local image, audio, video or document and get a URL to pass to any model
atlas_chat	Chat with LLM models — endpoint and request format are picked automatically per model (OpenAI chat/responses, Anthropic messages, native Gemini); supports image/video/audio input
atlas_get_prediction	Check status and result of a generation task — media URLs, transcripts, lyrics, cover art and cost
atlas_list_predictions	Browse past generation tasks — recover a lost prediction ID or review earlier results
atlas_get_balance	Get the account balance and credit summary for your API key
atlas_get_model_usage	Get daily model usage (requests, tokens, image/video counts) over a date range
atlas_get_model_costs	Get daily model cost (spend) buckets over a date range

Usage Examples

Search for models

"Search Atlas Cloud for video generation models"

Your AI assistant will use atlas_search_docs or atlas_list_models to find relevant models.

Generate an image

"Generate an image of a cat in space using Seedream"

The assistant will:

1. Use atlas_list_models to find Seedream image models
2. Use atlas_get_model_info to get the model's parameters
3. Use atlas_generate_image with the correct parameters

Generate a video

"Create a video of a rocket launch using Kling v3"

The assistant will:

1. Find the Kling video model
2. Get its schema to understand required parameters
3. Use atlas_generate_video with appropriate parameters

Upload a local image for editing or video generation

"Edit this image /Users/me/photos/cat.jpg to add a hat"

The assistant will:

1. Use atlas_upload_media to upload the local file and get a URL
2. Find an image-editing model
3. Use atlas_generate_image with the uploaded URL

Note: Uploaded files are for temporary use with Atlas Cloud generation tasks only. Files may be cleaned up periodically. Do not use this as permanent file hosting — abuse may result in API key suspension.

Generate speech (TTS)

"Read this sentence aloud with Seed Audio: Welcome to Atlas Cloud"

The assistant will:

1. Use atlas_list_models with type="Audio" to find the TTS model
2. Use atlas_generate_audio with the text to synthesize
3. Use atlas_get_prediction to retrieve the generated audio URL

Generate music

"Make a 30-second upbeat synthwave track for my product demo with Suno"

Music models (Suno Chirp, MiniMax Music) are Audio-type models, so the assistant uses atlas_generate_audio with a song description (and optionally lyrics), then retrieves the audio URL via atlas_get_prediction.

Transcribe audio (speech-to-text)

"Transcribe this interview recording: https://example.com/interview.mp3"

The assistant uses atlas_transcribe_audio with a speech-to-text model (e.g., bytedance/seed-asr-2.0) and the audio_url, then retrieves the transcript via atlas_get_prediction. For local files, it first calls atlas_upload_media to get a URL.

Generate a 3D model

"Turn this product photo into a 3D model with Hunyuan 3D"

3D models are Image-type models, so the assistant uses atlas_generate_image with the image parameter and retrieves a GLB/OBJ/USDZ file via atlas_get_prediction.

Chat with an LLM

"Ask Qwen to explain quantum computing"

The assistant will use atlas_chat with the Qwen model. For a model that accepts images or video, attach them to the message and they are converted to that model's protocol automatically.

Check a call before paying for it

"Show me exactly what you'd send to Kling for this, don't run it yet"

Every generation tool takes dry_run: true — it resolves the model, builds the request, validates it against the model's schema, prints the exact JSON body, and stops. Nothing is submitted and nothing is billed. Useful when a media URL could land on more than one input field, or when a keyword could match several models.

Find an earlier generation

"What did I generate yesterday? Give me the video link again"

The assistant uses atlas_list_predictions to list recent tasks and atlas_get_prediction for the full result. This is also how to recover a task whose prediction ID was lost — the job keeps running (and is billed) even if the submitting call timed out.

Check balance and usage

"How much Atlas Cloud credit do I have left, and what did I spend this month?"

The assistant will use atlas_get_balance for the current balance and atlas_get_model_costs for the spend breakdown.

Development

# Install dependencies
npm install

# Build
npm run build

# Run in development mode
npm run dev

More Atlas Cloud Tools

• 🧰 Want to use it from the terminal? → atlascloud-cli
• 🤖 Want to use it in Claude Code / Cursor? → Atlas Cloud MCP Server
• 🎬 Want it as a Claude Code / Codex / Gemini CLI Skill? → atlas-cloud-skills
• 🎨 ComfyUI nodes → atlascloud_comfyui
• 🔁 n8n nodes → n8n-nodes-atlascloud
• 💬 Join our Discord → discord.gg/MWmMr4q9es
• 🌐 Website → atlascloud.ai

License

MIT