Quickstart
From account to first call. About 5 minutes.
Step 1: Get an API Key
Whether you're wiring up a tool or writing code, it starts with these three things.
- Sign in — use the account GSS issued you at console. Enterprise accounts are provisioned by GSS; there is no self-signup.
- Check your quota — open Quota & API Keys and confirm your account or department has quota available. Calls fail outright when quota is empty.
- Create an API Key — create and copy the key on that same page. It starts with
sk-. Save it securely; it can be copied again from this page when its state and your permissions allow it.
If you need to create departments, invite members and allocate quota first, see Get an API Key.
Step 2: Note the Base URL
Every request goes to the same entry point:
https://api.smartwan.com
https://api.tokenroute.com is this platform's official API domain and is not yet active. Until the enablement announcement, use api.smartwan.com. Both will then serve in parallel, and your API Key will keep working unchanged.
These are complete request URLs. For SDK and client Base URL configuration, see Authentication:
| What you're calling | Full endpoint |
|---|---|
| OpenAI-compatible chat | https://api.smartwan.com/v1/chat/completions |
| Anthropic Messages | https://api.smartwan.com/v1/messages |
| Google Gemini | https://api.smartwan.com/v1beta/models/{model}:generateContent |
| Image generation | https://api.smartwan.com/v1/images/generations |
In client tools you'll usually enter https://api.smartwan.com or https://api.smartwan.com/v1 — follow Authentication. Some tools append /v1 automatically; duplicating it can cause requests to fail.
Step 3: Pick your path
I want to use a tool
Desktop clients like Chatbox and Cherry Studio, or CLI tools like Claude Code and Codex CLI.
- Pick one from All Tools and fill in API Key, Base URL and model name as that page describes
- Send a "hello"
- A reply means you're connected
I want to call it from code
Run these commands in Bash / Git Bash and replace the placeholder key. For Windows PowerShell and SDK examples, see First Call.
curl https://api.smartwan.com/v1/chat/completions \
-H "Authorization: Bearer ***" \
-H "Content-Type: application/json" \
-d '{
"model": "deepseek-v4-flash",
"messages": [{"role": "user", "content": "Hello"}]
}'
If the response contains the model's reply text, you're through. For Python, Node.js and other protocols, see Your First Call.
Verify your key first
When you're unsure whether the key or the network is the problem, request the model list — it's the lightest probe and consumes no quota:
curl https://api.smartwan.com/v1/models \
-H "Authorization: Bearer ***"
A returned model list proves this catalog request passed. Verify inference separately and use error details to diagnose failures:
| Response | What to investigate | What to do |
|---|---|---|
401 | Invalid key or malformed header | Check for a missing Bearer prefix, or a deleted key |
403 | No permission or quota exhausted | Check your balance in Quota & API Keys |
404 | Wrong endpoint | Check for a duplicated or missing /v1 |
| Connection timeout | Network unreachable | Confirm you can reach api.smartwan.com |
Which model to pick
Model names must match Models exactly — case and hyphens included. Copy deepseek-v4-flash exactly rather than changing it to DeepSeek-V4-Flash; error status depends on the route.
If you're unsure, start with deepseek-v4-flash: fast, and good enough for everyday chat and coding tasks. For heavier reasoning, switch to deepseek-v4-pro, claude-sonnet-5 or gpt-5.5.
Next steps
- Get an API Key — the full flow for admins: departments and quota allocation
- All Tools — how to connect each client and CLI tool
- Your First Call — complete examples for all four protocols
- FAQ — start here when you hit an error