OneProvider

OneProvider is a unified AI API gateway designed to let developers access mainstream AI models such as Claude, GPT, Gemini, and DeepSeek through a single API key and pay-as-you-go balance. It provides Anthropic- and OpenAI-compatible API endpoints, making it easy for users to switch models across different clients without changing their workflow.

Purpose

OneProvider is mainly aimed at developers, for directly calling multiple AI models in code, editors, or AI agents, without needing a separate website subscription for each model. It is suitable for production environments that need flexible use of multiple models, cost control, and simplified API management.

Main capabilities

• Unified access: A single API key can connect to models such as Claude, GPT, Gemini, and DeepSeek through familiar Anthropic- and OpenAI-compatible API calls.
• Pay-as-you-go: It uses a prepaid balance system. After users top up, billing is based on each model's per-token rate, with no automatic charges, and the balance does not expire. When the balance is exhausted, requests return a 429 "Insufficient balance" error until topped up.
• Model catalog: Provides a list of models in the current production catalog, including Claude Opus 5, Sonnet 5.5, Haiku 4.5, and others, as well as GPT, Gemini, DeepSeek, and others, with context size and per-million-token rates (including cache rates) noted.
• Compatibility: Supports Anthropic native endpoints and OpenAI-compatible endpoints. Claude Code and the Anthropic SDK use https://api.oneprovider.dev，OpenAI compatible tools use https://api.oneprovider.dev/v1。
• Management tools: Purchase, top up, issue keys, and view balance, usage, and activity history via the Telegram bot or web console (app.oneprovider.dev).
• Security and privacy: The service is operated by OneProvider Ltd. (Cyprus), with the gateway hosted in Frankfurt and using TLS 1.3. The policy states that prompt text is not stored. The terms, refund, and privacy policies are public, the status page shows real-time uptime, and every request is recorded in the usage log.

How to use

1. Register and top up your balance in the Telegram bot or web console.
2. Issue an API key in the console.
3. Configure the key and endpoint in a supported client (such as Claude Code, Cursor, n8n, etc.).
4. For Claude Code, after installation set the environment variable ONEPROVIDER_KEY to your key.
5. To switch models, simply change the model name in a compatible client.
6. The documentation provides setup guides for each client.

Use cases

• Developers need to flexibly switch between multiple AI models while maintaining a unified workflow.
• Want to control costs with pay-as-you-go pricing and avoid separate subscriptions for each model.
• Use tools such as Claude Code, Cursor, and n8n that require a gateway compatible with the Anthropic or OpenAI API.
• Need an alternative to gateways such as OpenRouter, and the tool supports a custom base URL.

Limitations

• Service availability, the specific model list, and rates may change; refer to the pricing page for the latest information.
• The balance is prepaid, and once exhausted it must be topped up manually.
• Prompt text is not stored, but other service data boundaries should be checked in the security description.
• Some features (such as cached rates) are available only on supported models.
• Specific performance metrics (such as request volume and savings percentage) come from official materials, and actual experience may vary depending on usage.