Setup & Connectivity
The connector starts but my requests return 503
The connector starts but my requests return 503
A 503 means the Tokios gateway can’t reach an active connector tunnel for the requested model. Check that:
- The connector process is still running (it should print “Tunnel established” at startup)
- The model name in your API request matches exactly what you registered in the dashboard
- Your
tunnel_tokenintokios.jsonmatches the token shown in the dashboard
Do I need to open any firewall ports?
Do I need to open any firewall ports?
No. The connector only dials outbound to
api.tokios.com over WebSockets (port 443). Your firewall needs no changes. Nothing on your machine listens for inbound connections.Can I run the connector on a different machine than my model?
Can I run the connector on a different machine than my model?
Yes. The
upstream URL can point to any host reachable from the machine running the connector — not just localhost. For example, http://192.168.1.50:11434 works if that machine runs Ollama and the connector can reach it on your LAN.API Compatibility
Which OpenAI API features does Tokios support?
Which OpenAI API features does Tokios support?
Tokios supports the core chat completions surface: messages, system prompts, streaming, temperature, and max_tokens. Features that depend on OpenAI-specific infrastructure (like Assistants, Files, or Fine-tuning) are not available — those require OpenAI’s backend.
Can I use streaming responses?
Can I use streaming responses?
Yes. Set
"stream": true in your request body. Tokios will return server-sent events in the standard OpenAI/Anthropic streaming format. Streaming is supported on all three endpoints.My client library sends requests to /v1 — will it work?
My client library sends requests to /v1 — will it work?
Yes. When configuring SDK clients, set the base URL to
https://api.tokios.com/v1 (with the /v1 suffix). When using environment variables like OPENAI_BASE_URL, set it to https://api.tokios.com (without /v1 — the SDK appends it automatically). Check your specific client’s documentation to see which format it expects.Local Models
Does Tokios store my prompts or model responses?
Does Tokios store my prompts or model responses?
No. Tokios routes requests through the tunnel to your local model and streams the response back. Your prompts and completions are not stored on Tokios infrastructure.
Which model backends are supported?
Which model backends are supported?
Any model server that exposes an HTTP API works: Ollama, llama.cpp server, vLLM, and LM Studio are the most common. Set the
upstream field in tokios.json to the server’s local URL.Can I use multiple models at once?
Can I use multiple models at once?
Yes. Run one connector instance per model, each with its own
tunnel_token and upstream pointing to a different model server. Register each connector as a separate deployment in the dashboard with a different name.Account & Billing
Is Tokios free?
Is Tokios free?
Yes, the community tier is free with no credit card required. You get full access to the gateway features to run and evaluate Tokios with your own hardware.
How do I delete an API key?
How do I delete an API key?
Go to your dashboard, navigate to API Keys, find the key you want to remove, and click Revoke. The key becomes invalid immediately.