Smart Routing
Automatically route requests to the best model based on cost, latency, quality, and capability. Fall back gracefully on errors.
Prism connects your stack to 350+ AI models across every modality — with intelligent routing, guardrails, RAG, and full observability. One API. Enterprise-grade.
// Intelligent routing across 350+ models
import { Prism } from '@prismgateway/sdk'
const prism = new Prism({
apiKey: process.env.PRISM_API_KEY
})
// Route to the best model automatically
const response = await prism.chat({
model: 'auto',
messages: [{
role: 'user',
content: 'Summarize this document'
}]
})
// ✓ Routed to claude-3-7-sonnet (best cost/quality)
// ✓ Guardrails applied — 0 PII detected
// ✓ Logged to observability dashboardTrusted by teams building the next generation of AI products
Each provider has a different SDK, auth method, and error format
No unified billing or cost visibility across Claude, GPT, Gemini
Zero observability — you don't know what's failing or why
Compliance teams can't audit AI requests without custom tooling
One OpenAI-compatible API for every model and modality
Unified billing dashboard with per-model cost breakdown
Full request tracing, latency analytics, and anomaly detection
Built-in guardrails, PII detection, and audit logs out of the box
Built for engineering teams that need reliability, control, and visibility — without sacrificing developer experience.
Automatically route requests to the best model based on cost, latency, quality, and capability. Fall back gracefully on errors.
Enforce content policies, detect PII, block prompt injections, and ensure compliance — before responses reach your users.
Text, images, audio, video, and embeddings — unified through one API. Switch modalities without rewriting your integration.
Connect your vector stores, knowledge bases, and document pipelines. Built-in chunking, embedding, and hybrid retrieval.
First-class Model Context Protocol support. Standardize how your agents and tools communicate across any model provider.
Full request tracing, cost analytics, latency percentiles, and anomaly detection. Know exactly what your AI spend is doing.
Prism is fully OpenAI-compatible. Change one line — baseURL — and you're done. No new SDK to learn, no migration cost.
Works with any OpenAI SDK (Python, TypeScript, Go, Ruby)
Zero code changes beyond one environment variable
Same request/response format you already know
Streaming, function calling, and embeddings all supported
// Drop-in replacement — just change baseURL
import OpenAI from 'openai'
const client = new OpenAI({
baseURL: 'https://api.prismgateway.io/v1',
apiKey: process.env.PRISM_API_KEY,
})
const response = await client.chat.completions.create({
model: 'auto', // Prism picks the best model
messages: [{ role: 'user', content: 'Summarize this doc' }]
})
// ✓ Routed → claude-3-7-sonnet ✓ Guardrails applied ✓ Cost: $0.002Start free. Scale as you grow. No hidden fees.
For individuals and side projects
No credit card required
For growing teams shipping AI products
For organizations with advanced needs
Free tier includes 5M tokens/month. No credit card required. Production-ready in minutes.