DEV Community

GPU-Bridge
GPU-Bridge

Posted on

GPU-Bridge Plugin for ElizaOS: Give Your Agents GPU Superpowers

#elizaos #ai #agents #gpu

GPU-Bridge Plugin for ElizaOS: Give Your Agents GPU Superpowers

ElizaOS agents are powerful. But when they need to run LLM inference, generate images, or transcribe audio, they typically hit the same problem: they can't hold a credit card.

The GPU-Bridge plugin solves this. It gives your ElizaOS agents access to 30 GPU-powered AI services — and they can pay autonomously using USDC on Base L2, with no human approval required.

Install 

npm install @gpubridge/plugin-gpubridge
Enter fullscreen mode Exit fullscreen mode

Get an API key at gpubridge.xyz, or skip it entirely for x402 autonomous mode.

Register in your agent 

import { gpuBridgePlugin } from '@gpubridge/plugin-gpubridge';

const agent = new AgentRuntime({
    plugins: [gpuBridgePlugin],
    // ...
});
Enter fullscreen mode Exit fullscreen mode

Set your API key in the environment:

GPUBRIDGE_API_KEY=gpub_...
Enter fullscreen mode Exit fullscreen mode

Or leave it unset for x402 autonomous payments — your agent pays with USDC from its own wallet, no humans needed.

What your agent can do

The plugin registers four ElizaOS model handlers:

TEXT_SMALL / TEXT_LARGE — LLM Inference

Your agent uses llm-4090 (Llama 3.3 70B) for text generation:

// ElizaOS calls this automatically for text tasks
const response = await runtime.generateText({ prompt: "Summarize this document..." });
Enter fullscreen mode Exit fullscreen mode

TEXT_EMBEDDING — Semantic Embeddings

1024-dimensional embeddings for semantic search and memory:

const vector = await runtime.embed({ text: "find similar content" });
Enter fullscreen mode Exit fullscreen mode

IMAGE — Image Generation

Generate images directly from your agent's reasoning:

const imageUrl = await runtime.generateImage({ prompt: "a cyberpunk city at sunset" });
Enter fullscreen mode Exit fullscreen mode

Direct service calls

The plugin also exports a client for direct access to all 30 services:

import { GPUBridgeClient } from '@gpubridge/plugin-gpubridge/client';

const client = new GPUBridgeClient({ apiKey: 'gpub_...' });

// Transcribe audio
const transcript = await client.transcribeAudio('https://example.com/audio.mp3');

// Generate image
const imageUrl = await client.generateImage('a robot reading a book');

// LLM inference
const text = await client.generateText('What is the capital of France?');
Enter fullscreen mode Exit fullscreen mode

x402 autonomous payments

The killer feature: your ElizaOS agent can pay for GPU compute without human intervention.

When no API key is set, GPU-Bridge returns an HTTP 402 with payment details. Your agent pays with USDC on Base L2 and the service executes. No credit card, no signup, no approval queue.

// No API key — agent pays autonomously
const agent = new AgentRuntime({
    plugins: [gpuBridgePlugin],
    // GPUBRIDGE_API_KEY not set
});
Enter fullscreen mode Exit fullscreen mode

This works because of x402 — a draft HTTP extension for machine-native micropayments. GPU-Bridge is one of the first production APIs to implement it.

All 30 services

GPU-Bridge exposes more than just LLM and images. Via the client:

Service Type
llm-4090 LLM inference (Llama 70B)
embedding-l4 Text embeddings
image-4090 Image generation
whisper-l4 Audio transcription
tts-l4 Text-to-speech
rerank Semantic reranking
pdf-parse PDF extraction
nsfw-detect Content moderation
video-enhance Video upscaling
+ 21 more See catalog

Full catalog: curl https://api.gpubridge.xyz/catalog

Links

Built for the agentic era: one endpoint, 30 services, pay with crypto. Follow @gpubridge for more.

Top comments (0)