Skip to main content
Toolken is a managed LLM gateway. Here are answers to the most common questions about how it works.
A managed LLM gateway on the edge. Point your client at one base URL and every call is attributed by feature, customer, and more, with rate limits enforced automatically. You bring your own provider keys.
No. Change one base URL to https://gateway.toolken.ai/v1 and add an X-Toolken-Key header. The gateway is OpenAI-compatible, so the rest of your code stays the same.
13 native providers are supported: OpenAI, Anthropic, Groq, Mistral, DeepSeek, Together, MiniMax, Gemini, xAI, OpenRouter, Cerebras, Fireworks, and Perplexity. Any OpenAI-compatible endpoint also works. OpenRouter unlocks a large additional catalog. Anthropic uses its own /v1/messages endpoint; Gemini is reached through the OpenAI-compatible endpoint.
Embeddings work and are cost-tracked like chat. Image and audio requests pass through to your provider and work today, but are not yet counted toward your cost totals. Cost tracking for them is coming soon.
A single hop at the edge. The gateway runs on Cloudflare’s edge network, close to your client, so the added latency is a small fraction of the total round trip to your provider.
Rate limits fail open, so a brief disruption lets requests through rather than blocking them. Authentication fails closed, so requests are never forwarded without a valid key.
No. Toolken is a managed gateway that runs on the edge. There is nothing for you to host.
Email [email protected] and the team will get back to you.
Still have questions? Email [email protected].