OpenRouter

What it does

OpenRouter takes standard HTTP requests in an OpenAI-style chat completions format, plus a model name and your API key, and returns model responses from whichever provider serves that model. The catalog covers 500+ models across 80+ providers, and the site describes generating text, images, video and audio through one interface. Responses can be streamed, and the API supports tool calling, structured JSON output and plugins such as PDF parsing, where the underlying model allows.

Developers swap their base URL and model slug in existing OpenAI-compatible code, or use the TypeScript and Python client SDKs or the Agent SDK for tool-using agent loops. What sets it apart is routing: it falls back to other providers when one errors or rate-limits, picks cost-effective options, and lets you set data policies on which providers see your prompts. A remote MCP server lets coding assistants look up live model, pricing and credit data.

It suits developers and teams who want several models behind one billing account without managing separate provider accounts. You need an account and an API key, and you buy credits; the free tier is capped at 50 requests a day on free models. This is an API service rather than a chat app, so you must write or configure code to use it. Latency varies by model and provider.

What it can do

• Single OpenAI-compatible chat completions endpoint for 500+ models
• Automatic fallback to other providers on errors or rate limits
• Streaming, tool calling and structured JSON outputs
• Data-policy-based routing and Zero Data Retention routing
• Workspaces with per-key credit caps for separating environments
• EU and US in-region routing on Business and Enterprise

Works with

• OpenAI SDK (drop-in with changed base URL)
• OpenRouter MCP server for coding assistants

Example workflows

Migrate an OpenAI-based app

Inputs: existing OpenAI-compatible code and an OpenRouter API key. Steps: change the base URL to OpenRouter, set a model slug from the catalog, send a test request. Output: chat completions served by the chosen model, with fallback to other providers on errors.

Extract structured data from documents

Inputs: a PDF and a JSON schema. Steps: choose a model supporting structured outputs, set response_format with the schema, enable the PDF plugin if needed. Output: JSON rows matching the schema, to be validated by your own checks.

Before you choose

• Credit purchases carry a platform fee (5.5% Standard, 8% Business); inference itself is billed at provider list price.
• Free models are limited to 20 requests per minute and 50 per day, rising to 1,000 per day after buying at least $10 of credits.
• Parameters a model does not support are ignored, so behavior can differ between models.

Listed under

• Coding
• AI models
• API
• API gateway
• Developers
• Fallback routing
• Fallbacks

Get started

Buy credits to use paid models

1. Create an account at openrouter.ai and buy credits (or use free models).
2. Create an API key and store it as OPENROUTER_API_KEY.
3. Send a request to /api/v1/chat/completions or install @openrouter/sdk or the openrouter Python package.

Inference is billed per input/output token at each provider's list price. OpenRouter's fee applies when buying credits: 5.5% Standard, 8% Business. Enterprise is set by contract. BYOK includes $25,000/month list-price inference, then 5%.

Source: openrouter.ai/pricing, checked 29 Oct 2026. Prices can change; check before you buy.

Sources for this overview

Product pages, documentation, and pricing checked 29 Oct 2026. Each source supports part of the overview; availability and prices can change.

• OpenRouter Quickstart Guide ↗openrouter.ai
• OpenRouter ↗openrouter.ai
• Pricing | OpenRouter ↗openrouter.ai
• OpenRouter API Reference - Complete Documentation ↗openrouter.ai

Company website

openrouter.ai: product, pricing and policy pages

YC directory mirror

Batch and company name

Web search

Confirms the official site

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 29 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.