Radox AI offers a way to try connecting Claude Desktop to a third-party API gateway rather than accessing Anthropic directly. The service advertises support for Claude, GPT, Gemini, DeepSeek and other models. Its homepage mentions a free allowance of 0.5 million tokens per month with no credit card required, but prospective users should check the current limit before signing up.

What Radox AI offers — and why its limits need checking

Radox AI presents itself as a unified API for dozens of models. Its homepage promotes a $4.50-per-month Starter plan with 100 million tokens and an Unlimited plan for $50. However, the pricing page describes the Free plan as including 10 million tokens per month rather than 0.5 million. Because the published figures conflict, the exact allowance should be checked in the dashboard or with support before use.

This is not an Anthropic promotion or a free Claude Pro or Max subscription: requests are routed through an external API provider. A model name appearing in a provider's list does not establish that model's provenance or actual availability; those details require separate verification.

How to configure Claude Desktop

1. Download Claude Desktop from the official website. You do not have to complete the usual sign-in flow first if you are configuring a third-party inference provider.

2. Open Help → Troubleshooting → Enable Developer Mode. Then select Developer → Configure Third-Party Inference. Depending on the app version, a restart may be required.

3. Create an account with Radox AI and generate an API key. Keep the key private. The example uses https://radoxai.com as the gateway address; verify the current endpoint and supported protocol against Radox AI's documentation.

4. In Connection, set Inference provider to Gateway, Credential kind to Static API key and Gateway auth scheme to Bearer. Enter the Gateway base URL without the /v1 suffix and paste the key into Gateway API key.

5. Select Test connection. If the client shows Connected and discovers models, apply the configuration with Apply changes followed by Save & Restart. A successful connection confirms a technical response from the gateway, not the quality or provenance of every listed model.

What to keep in mind

A free allowance may be enough for short experiments, but larger models and longer context windows can consume tokens quickly. Before paying for a larger package, review the current terms, rate limits, overage charges and refund conditions.

Security matters too: prompts, code snippets and documents may pass through a third-party operator's infrastructure. Avoid sending sensitive information until you have reviewed the provider's data-handling policy and terms of service.

Claude Desktop needs a gateway that supports the protocol it expects. Advertising an OpenAI-compatible API alone does not guarantee that every Claude Desktop feature will work. Start with a harmless test request.

Sources

Radox AI — official website

Radox AI — pricing

Cloudflare guide to configuring a third-party gateway in Claude Desktop

Radox AI offers a way to try connecting Claude Desktop to a third-party API gateway rather than accessing Anthropic directly. The service advertises support for Claude, GPT, Gemini, DeepSeek and other models. Its homepage mentions a free allowance of 0.5 million tokens per month with no credit card required, but prospective users should check the current limit before signing up.

What Radox AI offers — and why its limits need checking

Radox AI presents itself as a unified API for dozens of models. Its homepage promotes a $4.50-per-month Starter plan with 100 million tokens and an Unlimited plan for $50. However, the pricing page describes the Free plan as including 10 million tokens per month rather than 0.5 million. Because the published figures conflict, the exact allowance should be checked in the dashboard or with support before use.

This is not an Anthropic promotion or a free Claude Pro or Max subscription: requests are routed through an external API provider. A model name appearing in a provider's list does not establish that model's provenance or actual availability; those details require separate verification.

How to configure Claude Desktop

1. Download Claude Desktop from the official website. You do not have to complete the usual sign-in flow first if you are configuring a third-party inference provider.

2. Open Help → Troubleshooting → Enable Developer Mode. Then select Developer → Configure Third-Party Inference. Depending on the app version, a restart may be required.

3. Create an account with Radox AI and generate an API key. Keep the key private. The example uses https://radoxai.com as the gateway address; verify the current endpoint and supported protocol against Radox AI's documentation.

4. In Connection, set Inference provider to Gateway, Credential kind to Static API key and Gateway auth scheme to Bearer. Enter the Gateway base URL without the /v1 suffix and paste the key into Gateway API key.

5. Select Test connection. If the client shows Connected and discovers models, apply the configuration with Apply changes followed by Save & Restart. A successful connection confirms a technical response from the gateway, not the quality or provenance of every listed model.

What to keep in mind

A free allowance may be enough for short experiments, but larger models and longer context windows can consume tokens quickly. Before paying for a larger package, review the current terms, rate limits, overage charges and refund conditions.

Security matters too: prompts, code snippets and documents may pass through a third-party operator's infrastructure. Avoid sending sensitive information until you have reviewed the provider's data-handling policy and terms of service.

Claude Desktop needs a gateway that supports the protocol it expects. Advertising an OpenAI-compatible API alone does not guarantee that every Claude Desktop feature will work. Start with a harmless test request.

Sources

Radox AI — official website

Radox AI — pricing

Cloudflare guide to configuring a third-party gateway in Claude Desktop

Radox AI offers a way to try connecting Claude Desktop to a third-party API gateway rather than accessing Anthropic directly. The service advertises support for Claude, GPT, Gemini, DeepSeek and other models. Its homepage mentions a free allowance of 0.5 million tokens per month with no credit card required, but prospective users should check the current limit before signing up.

What Radox AI offers — and why its limits need checking

Radox AI presents itself as a unified API for dozens of models. Its homepage promotes a $4.50-per-month Starter plan with 100 million tokens and an Unlimited plan for $50. However, the pricing page describes the Free plan as including 10 million tokens per month rather than 0.5 million. Because the published figures conflict, the exact allowance should be checked in the dashboard or with support before use.

This is not an Anthropic promotion or a free Claude Pro or Max subscription: requests are routed through an external API provider. A model name appearing in a provider's list does not establish that model's provenance or actual availability; those details require separate verification.

How to configure Claude Desktop

1. Download Claude Desktop from the official website. You do not have to complete the usual sign-in flow first if you are configuring a third-party inference provider.

2. Open Help → Troubleshooting → Enable Developer Mode. Then select Developer → Configure Third-Party Inference. Depending on the app version, a restart may be required.

3. Create an account with Radox AI and generate an API key. Keep the key private. The example uses https://radoxai.com as the gateway address; verify the current endpoint and supported protocol against Radox AI's documentation.

4. In Connection, set Inference provider to Gateway, Credential kind to Static API key and Gateway auth scheme to Bearer. Enter the Gateway base URL without the /v1 suffix and paste the key into Gateway API key.

5. Select Test connection. If the client shows Connected and discovers models, apply the configuration with Apply changes followed by Save & Restart. A successful connection confirms a technical response from the gateway, not the quality or provenance of every listed model.

What to keep in mind

A free allowance may be enough for short experiments, but larger models and longer context windows can consume tokens quickly. Before paying for a larger package, review the current terms, rate limits, overage charges and refund conditions.

Security matters too: prompts, code snippets and documents may pass through a third-party operator's infrastructure. Avoid sending sensitive information until you have reviewed the provider's data-handling policy and terms of service.

Claude Desktop needs a gateway that supports the protocol it expects. Advertising an OpenAI-compatible API alone does not guarantee that every Claude Desktop feature will work. Start with a harmless test request.

Sources

Radox AI — official website

Radox AI — pricing

Cloudflare guide to configuring a third-party gateway in Claude Desktop

Radox AI offers a way to try connecting Claude Desktop to a third-party API gateway rather than accessing Anthropic directly. The service advertises support for Claude, GPT, Gemini, DeepSeek and other models. Its homepage mentions a free allowance of 0.5 million tokens per month with no credit card required, but prospective users should check the current limit before signing up.

What Radox AI offers — and why its limits need checking

Radox AI presents itself as a unified API for dozens of models. Its homepage promotes a $4.50-per-month Starter plan with 100 million tokens and an Unlimited plan for $50. However, the pricing page describes the Free plan as including 10 million tokens per month rather than 0.5 million. Because the published figures conflict, the exact allowance should be checked in the dashboard or with support before use.

This is not an Anthropic promotion or a free Claude Pro or Max subscription: requests are routed through an external API provider. A model name appearing in a provider's list does not establish that model's provenance or actual availability; those details require separate verification.

How to configure Claude Desktop

1. Download Claude Desktop from the official website. You do not have to complete the usual sign-in flow first if you are configuring a third-party inference provider.

2. Open Help → Troubleshooting → Enable Developer Mode. Then select Developer → Configure Third-Party Inference. Depending on the app version, a restart may be required.

3. Create an account with Radox AI and generate an API key. Keep the key private. The example uses https://radoxai.com as the gateway address; verify the current endpoint and supported protocol against Radox AI's documentation.

4. In Connection, set Inference provider to Gateway, Credential kind to Static API key and Gateway auth scheme to Bearer. Enter the Gateway base URL without the /v1 suffix and paste the key into Gateway API key.

5. Select Test connection. If the client shows Connected and discovers models, apply the configuration with Apply changes followed by Save & Restart. A successful connection confirms a technical response from the gateway, not the quality or provenance of every listed model.

What to keep in mind

A free allowance may be enough for short experiments, but larger models and longer context windows can consume tokens quickly. Before paying for a larger package, review the current terms, rate limits, overage charges and refund conditions.

Security matters too: prompts, code snippets and documents may pass through a third-party operator's infrastructure. Avoid sending sensitive information until you have reviewed the provider's data-handling policy and terms of service.

Claude Desktop needs a gateway that supports the protocol it expects. Advertising an OpenAI-compatible API alone does not guarantee that every Claude Desktop feature will work. Start with a harmless test request.

Sources

Radox AI — official website

Radox AI — pricing

Cloudflare guide to configuring a third-party gateway in Claude Desktop