Integration overview
Put your TokenRoute API Key into the client, IDE or command-line tool you already use.
Use GSS Deploy (Windows one-click setup) for ChatGPT, Claude Code, Hermes, or WorkBuddy. Install it, select a tool locally, enter your API Key, and it writes the configuration for you.
Which tool
| What you want | Use this | Difficulty |
|---|---|---|
| Everyday chat, quick model trials | Chatbox | Low |
| Desktop client managing several models | Cherry Studio | Low |
| Writing code in an IDE | VS Code / Cline | Medium |
| Letting AI edit code from the terminal | Codex CLI | Medium |
| A Claude-style CLI | Claude Code | Medium |
| Adding custom models in WorkBuddy | WorkBuddy | Medium |
| Agent workflows with multiple profiles and toolsets | Hermes | High |
If you have no preference, start with Chatbox — install, fill three fields, done. It's the fastest way to confirm your key works.
Before you start
Four prerequisites, all required:
| Requirement | Notes |
|---|---|
| API Key | Created in the console — see Get an API Key |
| Available quota | Your account or department must have balance, or calls fail immediately |
| Model name | Copy from Models — case-sensitive |
| Network reachability | Your machine can reach api.smartwan.com |
Common configuration fields
Tools name these fields differently, but they all want the same few things:
| Field | What to enter |
|---|---|
| API Key | The sk- key created in the console |
| Base URL | https://api.smartwan.com or https://api.smartwan.com/v1 — follow the tool's page |
| Model | e.g. deepseek-v4-flash, claude-sonnet-5, gpt-5.5 |
| API type | OpenAI Compatible or Anthropic, depending on what the tool supports |
This is the most common trap. Some tools append /v1 automatically; add it yourself as well and you get /v1/v1/chat/completions and a 404.
Every tool page states the recommended value — follow it.
Verifying your setup
First click "fetch models" / "refresh model list" in the tool — if the list loads, both key and network are working.
Then start a new conversation and send:
Hello, please reply with: setup successful
A normal reply means you're connected.
If it fails, check in this order:
- Was the API Key copied in full (no characters lost at either end)?
- Is
/v1in the Base URL duplicated or missing? - Does the model name match the model list exactly?
- Does the account still have quota?
- Protocol and client authentication — Messages accepts
x-api-keyor Bearer; follow Authentication
Still stuck? See the FAQ.
Verification scope
| Scope | What was confirmed |
|---|---|
| Python SDKs | Minimal live calls with OpenAI 3.17.0, Anthropic 1.7.0 and Google GenAI 2.24.0 |
| CLIs | Minimal live calls with Codex CLI 0.155.1 + gpt-5.5, and Claude Code 2.1.278 + claude-haiku-4-5 |
| Hermes | Provider matching, configuration parsing and stable tool-management commands; the complete Agent / IM flow needs separate validation |
| GUI clients, IDEs and installer | Configuration instructions do not establish complete flow verification for your version, OS and account |
The live calls above were run on 2026-09-22. After upgrading a client or changing models, repeat a minimal call before restoring tools and advanced parameters.
Platform notes
| OS | Notes |
|---|---|
| Windows | Prefer the tool's own GUI configuration. For CLI tools, note that PowerShell and CMD set environment variables differently |
| macOS | Homebrew works well for installing Node.js and similar dependencies; config files usually live in ~/.config, ~/.claude, ~/.codex |
| Linux | Shell-based configuration mostly; mind environment variable scope and file permissions |