v1.0 TypeScript Pythonsoon Rustsoon

One API for every
LLM provider — that you own

No proxy to deploy, no take-rate, no second vendor. Your keys, your traffic, direct to OpenAI, Anthropic, Google, xAI, and OpenRouter. Zero runtime dependencies. Runs in the browser.

Get started Open Sandbox

quickstart.ts

import { complete } from '@combycode/llm-sdk';

const { text } = await complete({
  model: 'anthropic/claude-opus-4.8',
  apiKey: process.env.ANTHROPIC_KEY,
  prompt: 'Write a haiku about the sea.',
});

console.log(text);

Everything you need. Nothing you don't.

Built for teams who want unification as code they own — not another hop in the call chain.

◎

One API, every provider

Switch models with one string. We absorb OpenAI Responses, Anthropic Messages, Google generateContent — one interface behind all of it.

⬡

Zero dependencies

dependencies: {}. No transitive CVE surface, tiny install, fast cold start. Works on the edge, in serverless, anywhere.

◐

Runs in the browser

True cross-env: Node, Bun, and the browser. Lazy node: loaders, browser bundle shipped. Call LLMs directly from the client.

◈

Cost tracking built in

Bundled versioned model catalog + provider usage. Per-scope spend, budgets, alerts — no server to deploy, no gateway take-rate.

◉

Observability built in

OTel-style traces, metrics, logs — emitted in-process, no @opentelemetry dep, no proxy hop. Point at Langfuse or read the built-ins.

◫

Composable layers

Four optional layers: Network → Client → Agent → Server. Use a one-liner, or wire the pieces. No abstraction tax.

Official SDKs vs Gateways vs ORXA

Official SDKs give you one provider and nothing else. Gateways give you unification as a paid hop your traffic passes through. We give you the same unification as code you own — direct to providers, free, with cost and observability built in, in any runtime.

See full comparison

	Official SDK	Gateway	ORXA
Multi-provider	no	yes	yes
Cost tracking	no	yes*	yes
Network hop	no	yes	no
Take-rate	no	often	no
Zero deps	no	n/a	yes
Runs in browser	~	no	yes

One API for every LLM provider — that you own