OpenRouterTeam/ai-sdk-provider

The OpenRouter provider for the Vercel AI SDK contains support for hundreds of models through the OpenRouter chat and completion APIs.

What it solves

This project provides a bridge between the Vercel AI SDK and OpenRouter, allowing developers to access over 300 large language models and various embedding models through a single, standardized interface. It eliminates the need to write custom integration code for every individual model provider available on OpenRouter.

How it works

It acts as a provider instance for the Vercel AI SDK. By importing the openrouter provider, developers can specify OpenRouter model identifiers (e.g., openai/gpt-4o) within the SDK's standard functions like generateText, streamText, and embed. It supports advanced configuration options such as passing extra request bodies, enabling Anthropic-specific prompt caching and beta features (like fine-grained tool streaming), and integrating with OpenRouter-specific plugins like Response Healing.

Who it’s for

JavaScript and TypeScript developers using the Vercel AI SDK who want to easily switch between or utilize a vast array of different LLMs and embedding models without changing their core application logic.

Highlights

  • Massive Model Access: Access to 300+ LLMs and multiple embedding models via one SDK.
  • Structured Output Support: Integration with the Response Healing plugin to automatically repair malformed JSON responses.
  • Advanced Model Control: Support for Anthropic prompt caching and fine-grained tool streaming for reduced latency in large JSON structures.
  • Detailed Usage Tracking: Built-in support for usage accounting, including cost and token breakdown, including BYOK (Bring Your Own Key) scenarios.
  • Debugging Tools: A debug mode that echoes the upstream request body to help troubleshoot API calls.

Related

  • Project
    vercel/ai

    A provider-agnostic TypeScript toolkit for building AI-powered applications and agents with a unified API for multiple LLM providers and integrated UI hooks.

  • Project
    simonw/llm-openrouter

    A plugin for the LLM CLI utility that enables access to a wide variety of AI models hosted by OpenRouter, including support for vision, structured outputs, and hosted server tools.

  • Project
    fruitbars/simple-one-api

    simple‑one‑api is an open‑source Go gateway that unifies multiple LLM providers behind OpenAI‑compatible endpoints, adds capacity‑aware key scheduling, four‑layer rate‑limiting, a built‑in React chat UI, a visual config console, and a Wails desktop client. It can be run as a single binary, Docker container, or desktop app, and stores its configuration in SQLite.

  • Project
    adaline/gateway

    A production-grade TypeScript SDK that provides a unified, local interface for calling over 300 LLMs across multiple providers.

  • Project
    ben-vargas/ai-sdk-provider-claude-code

    An unofficial npm package that bridges the Vercel AI SDK with Anthropic’s Claude Code models via the official Claude Agent SDK. It supports streaming, multi‑turn sessions, native structured outputs (Zod schemas), tool usage, image handling, and fine‑grained Agent SDK options. Multiple provider versions map to AI SDK v4‑v7, with the current stable line (4.x) requiring Node ≥ 22 and Zod ^4.1.8.