Theta EdgeCloud Launches Official MCP Server for AI Model Access

Today, we’re announcing the release of an official MCP server that brings Theta EdgeCloud’s 18+ AI models to any MCP-compatible client, including Claude Desktop, Cursor, Cline, Zed, and others. This integration allows users to generate images, transcribe audio, and run LLM inference directly from their AI assistant using natural language.

What is MCP?

Model Context Protocol (MCP) is an open standard created by Anthropic that lets AI assistants connect to external tools and services. Think of it as a universal plugin system for AI: one integration works across all MCP-compatible clients.

MCP solves a fundamental problem. Every AI assistant previously needed its own custom integration for each external service. If you wanted to connect five different AI tools to the same API, you’d need five separate integrations. MCP changes this by providing a standardized way for AI assistants to communicate with external services.

With our official MCP server, any MCP client gains access to Theta EdgeCloud’s full suite of on-demand AI models without requiring custom code or API integration work for each client.

Why an MCP Server?

One integration, many clients. Build once, use everywhere. The same MCP server works with Claude Desktop, Cursor, Cline, Zed, Sourcegraph Cody, and any future MCP-compatible tool. Instead of maintaining separate integrations for each AI assistant, developers get universal access through a single implementation.

Natural language interface. Instead of writing API calls, describe what you want: “Generate an image of a mountain sunset” or “Transcribe this audio file.” The AI assistant handles the API interaction, parameter formatting, and result parsing automatically.

Context-aware. Your AI assistant understands conversation history, can iterate on results, and chain operations together naturally. For example, you can ask it to generate an image, then refine the prompt based on the result, then upscale the final version. All in a single conversation without switching tools.

No code required. For quick tasks, skip the SDK entirely. Just ask. This is particularly valuable for prototyping, one-off tasks, or teams where not everyone writes code.

Practical example: A content creator working in Claude Desktop can describe the image they need, have Claude generate it using FLUX, review the result, request adjustments, upscale it with ESRGAN, and export the final file. All without leaving their conversation or touching any code.

Available Tools

The MCP server exposes four tools:

list_services — Discover all available AI models and their capabilities

infer — Run inference on any model (image generation, transcription, LLMs, etc.)

get_request_status — Check status of async requests

get_upload_url — Get presigned URLs for file uploads

Available Models

Speech and Audio

Whisper — Speech-to-text transcription (99+ languages) Kokoro — Text-to-speech synthesis Voice Cloning — Clone voices from samples

Image Generation

FLUX.1 Schnell — Fast, high-quality generation Stable Diffusion XL Turbo — Real-time generation Stable Diffusion Turbo Vision — Vision-guided generation

Large Language Models

Llama 3.1 70B — Meta’s powerful open LLM DeepSeek R1 — Advanced reasoning model

Image Processing

ESRGAN — AI upscaling Grounding Dino — Object detection BLIP — Visual question answering

Video

Stable Diffusion Video — Text-to-video Talking Head — Portrait animation

Full list: https://ondemand.thetaedgecloud.com/service/list

Installation

The MCP server runs locally on your machine and connects your AI assistant to Theta EdgeCloud’s API. Installation requires an access token from Theta EdgeCloud and adding the server configuration to your AI assistant’s settings.

Step 1: Get an Access Token

Go to the On-Demand Models page: https://www.thetaedgecloud.com/dashboard/ai/service/on-demand-model-apis Click on any model Go to the “Access Token” tab Create or copy your access token (The same token works for all models. You only need one.)

Step 2: Add the MCP Server

The installation varies by client. Here are the most common:

Claude Code:

claude mcp add tec-on-demand-api \

-e THETA_API_KEY=your-access-token \

— npx @thetalabs/on-demand-api-mcp

Claude Desktop / Cursor / Other clients:

Add to your MCP configuration file:

{

“mcpServers”: {

“tec-on-demand-api”: {

“command”: “npx”,

“args”: [“@thetalabs/on-demand-api-mcp”],

“env”: {

“THETA_API_KEY”: “your-access-token”

}

}

}

}

Step 3: Use It

Ask your AI assistant:

“List available services on Theta EdgeCloud”
“Generate an image of a cyberpunk cityscape”
“Transcribe this audio: https://example.com/audio.wav"

Example Usage

Image Generation:

“Generate an image of a cozy coffee shop, warm lighting, rainy day outside”

The assistant calls infer(service=”flux”, input={prompt: “…”}) and returns an image URL.

Audio Transcription:

“Transcribe this podcast: https://example.com/episode.mp3"

The assistant calls infer(service=”whisper”, input={audio_filename: “…”}) and returns transcribed text.

Model Discovery:

“What image models are available?”

The assistant calls list_services() and returns a formatted list of image generation models.

How This Differs from Direct API Use

While you can use Theta EdgeCloud’s API directly through SDKs or HTTP requests, the MCP server offers distinct advantages for interactive workflows:

Conversational refinement. Instead of writing new code to adjust parameters, describe what you want changed. “Make it darker” or “try a different style” work naturally.

Automatic format handling. The AI assistant manages file uploads, presigned URLs, and response parsing. You don’t need to handle these details manually.

Workflow composition. Chain multiple AI operations together through conversation. Generate an image, upscale it, then transcribe audio describing it. All without explicit orchestration code.

Lower barrier to entry. Non-developers can access EdgeCloud’s AI models through familiar chat interfaces without learning API syntax or authentication patterns.

For production applications, batch processing, or automated pipelines, direct API access remains the better choice. The MCP server excels at interactive exploration, prototyping, and ad-hoc tasks.

Sync and Async Support

For quick tasks, the MCP waits for results (up to 60 seconds):

python

infer(service=”flux”, input={prompt: “A sunset”}, wait=30)

For longer tasks, use async mode and poll. First, submit the request:

python

infer(service=”whisper”, input={audio_filename: “…”}, wait=0)

This returns a request_id. Then check status:

python

get_request_status(request_id=”infr_rqst_xxx”)

This returns the result when ready.

Powered by Decentralized Infrastructure

All inference runs on Theta EdgeCloud’s hybrid cloud-edge network, which dynamically routes requests across cloud providers, enterprise data centers, and community-operated nodes based on performance requirements and availability.

Lower latency. Requests route to the nearest available GPU with capacity, reducing round-trip time compared to fixed data center locations.

High availability. The distributed architecture eliminates single points of failure. If one node or region experiences issues, requests automatically route to available alternatives.

Competitive pricing. Distributed infrastructure reduces operational costs compared to traditional centralized cloud providers. Academic institutions like KAIST, Yonsei University, and NTU Singapore use EdgeCloud for AI research at less than 50% the cost of conventional cloud GPU providers.

The network currently spans over 30,000 distributed edge nodes combined with cloud partnerships, providing hundreds of PetaFLOPS of computing capacity. This infrastructure powers both the interactive MCP interface and production API workloads for research institutions and enterprises.

Links

npm: https://www.npmjs.com/package/@thetalabs/on-demand-api-mcp

GitHub: https://github.com/thetalabs/on-demand-api-mcp

On-Demand Models: https://www.thetaedgecloud.com/dashboard/ai/service/on-demand-model-apis

API Docs: https://docs.thetaedgecloud.com

Service List: https://ondemand.thetaedgecloud.com/service/list

Theta EdgeCloud provides decentralized AI infrastructure for developers and enterprises. Learn more at thetaedgecloud.com.

Theta EdgeCloud Launches Official MCP Server for AI Model Access was originally published in Theta Network on Medium, where people are continuing the conversation by highlighting and responding to this story.