Official-model API access
The service exposes the same official Anthropic and OpenAI model versions and capabilities according to the product page, rather than a modified or tuned variant.
derouter.ai
derouter.ai is a web API for Claude and GPT models with fixed, discounted token pricing, plus support for Claude Code, Codex CLI, and GPT Image 2.
Overview
derouter.ai is an API service for accessing Claude and GPT models through one account and one API key. The product page describes it as a drop-in replacement for the official Anthropic and OpenAI APIs, with the same model versions and capabilities, and no request rewriting or model swapping.
Its main purpose is to give developers and teams lower-cost access to these frontier models with fixed, per-token pricing. The site says Claude models are priced at about one quarter of official list price and GPT models from 10% of official list price, while still supporting standard OpenAI-compatible clients and CLI workflows such as Claude Code and Codex CLI.
Core features
Transparent token pricing
The pricing page lists flat, fixed per-token rates in USD for input, output, cache read, and cache write, so costs do not vary with network load or hidden multipliers.
Compatible API routing
The home page says the API works as a drop-in replacement for OpenAI and Anthropic integrations, which means existing clients can usually switch by changing the base URL.
Developer integration guides
The site documents support for Python, JavaScript, and curl in the integration guide, plus CLI-oriented setup notes for Claude Code and Codex CLI.
Reseller tooling and guidance
The documentation includes reseller guides for sub-keys, markups, client management, and portal setup, aimed at teams that want to resell API access.
Image generation pricing
The pricing page includes GPT Image 2 image generation pricing based on output size tiers, with the quality setting not affecting the billed price.
Common use cases
Swap into an existing application
Developers who already use OpenAI-compatible SDKs can switch endpoints by updating the base URL instead of rewriting their app around a new API format.
Use Claude Code or Codex CLI with alternate routing
Users of Claude Code and Codex CLI can route requests through the service using the documented environment variable or custom provider setup.
Control API costs more predictably
Teams looking to reduce model spend can use the fixed token pricing to estimate usage more predictably than with variable multipliers or load-based pricing.
Handle text and image workflows together
Builders who need image generation as well as text generation can keep both workloads under one account using the listed GPT Image 2 pricing.
Operate a reseller or managed-client setup
Resellers and agencies can create sub-keys, set markups, and manage clients using the documentation aimed at portal and reseller setup.
Pros and Cons
Pros
- Uses fixed, transparent pricing rather than variable multipliers or network-based rate changes.
- Supports standard OpenAI-compatible clients by changing the base URL, which lowers switching friction.
- Covers both text models and GPT Image 2 image generation on one platform.
- Provides documentation for direct integration, CLI workflows, reselling, and referrals.
- States that requests are forwarded verbatim without prompt rewriting or model swapping.
Cons
- Cursor is not supported at the moment, according to the FAQ.
- Some integration details are still only documented at a high level in the source material, so setup may require checking the docs for client-specific steps.
FAQ
What does derouter.ai provide?
derouter.ai provides Claude and GPT APIs with the same official model versions and capabilities, exposed through one account and one API key. The site describes it as a drop-in replacement for the OpenAI and Anthropic APIs, with no prompt rewriting or model swapping.
How does pricing work?
The pricing page shows fixed, per-token pricing billed in USD per 1M tokens for text models, with separate cache and image pricing. The site states that Claude models are priced at about 1/4 of the official list price and GPT models from 10% of official list price, with no hidden multipliers.
How do I connect my app or CLI to it?
The site says standard OpenAI SDKs and compatible clients work after changing the base URL. For Claude Code, it lists ANTHROPIC_BASE_URL=https://api.derouter.ai/proxy, and for Codex CLI it says to configure a custom provider as shown in the docs.
Where is it available and how do I pay?
The product page says it is available worldwide, uses email sign-up, and supports payment by crypto or bank card. It also states there is no KYC, no phone number, and no credit card required for signup.
Are there any limitations I should know about?
The site states Cursor is not supported at the moment. It also says requests are forwarded verbatim and that the only added element is a standard client identity header.
Quick Facts
- Category
- Developer Tool
- Primary use
- API access to Claude, GPT, and image generation models
- Platform
- Web API
- Compatibility
- OpenAI API-compatible clients and Anthropic-style workflows
- Availability
- Worldwide, according to the site
- Domain
- derouter.ai
derouter.ai Alternatives
AakarDev AI
AakarDev AI helps teams manage AI provider access, project setup, logs, and analytics in one dashboard. BYOK support included.
ByteAsk
ByteAsk is a terminal-first AI coding agent for C and C++ that edits repos and verifies changes with compilers, debuggers, sanitizers, and tests.
CreateOS Sandbox
CreateOS Sandbox is an isolated compute environment for running code and agent workloads in Firecracker micro-VMs with private networking and SDK, CLI, or MCP control.
hob
hob is an independent workspace for coding agents, with local control over sessions, terminals, history, routing, and follow-up work.
Ably Chat
Ably Chat is a chat API platform for custom realtime chat apps, with rooms, typing indicators, presence, reactions, message updates and usage-based pricing.
Manta AI
Manta AI is an autonomous web app testing tool that maps app behavior, catches regressions, and generates tests from a URL, no scripts or selectors needed.