Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.utoken.wgame-hk.com/llms.txt

Use this file to discover all available pages before exploring further.

​
1. Overview

Claude Code is Anthropic’s official terminal AI coding client. Through Utoken (Anthropic-compatible API) you get:

Multi-vendor access

Use the full Claude family, Kimi K2, GPT and more behind one endpoint

Dual-thread model

Heavy tasks on the main thread, lightweight ones on the haiku thread

No subscription

Pay per token — no Claude Pro required

Stable globally

Distributed nodes with automatic failover

​
2. Install Node.js (skip if installed)

Make sure Node.js is ≥ 18.0.
Download: https://nodejs.org/dist/v22.18.0/node-v22.18.0-x64.msi

​
Uninstall other proxy clients (optional)

If you previously installed another forked @anthropic-ai/claude-code, clean it up first: 
# Check installation locations
npm ls @anthropic-ai/claude-code
npm ls -g @anthropic-ai/claude-code

# Uninstall
npm uninstall @anthropic-ai/claude-code
npm uninstall -g @anthropic-ai/claude-code

​
3. Install Claude Code

1

Global install

npm install -g @anthropic-ai/claude-code
2

Verify install

claude --version

​
4. Get your Utoken key

1

Sign in

Go to the Utoken console.
2

Create a key

Open API Keys → create a new token.
3

Note the info

base_url = "https://utoken-test.uuzu.com"
api_key  = "sk-zbTYx*******************************fceTvn5"
Claude Code’s ANTHROPIC_BASE_URL does not need a /v1 suffix — use https://utoken-test.uuzu.com directly.

​
5. Configure Environment Variables

​
5.1 Basic example (single model)

export ANTHROPIC_BASE_URL="https://utoken-test.uuzu.com"
export ANTHROPIC_AUTH_TOKEN="sk-zbTYx*******************************fceTvn5"
export ANTHROPIC_MODEL="kimi-k2-250905"
cd your-project-folder
claude

​
5.2 Dual-thread architecture

By default Claude Code uses a dual-thread model architecture that saves cost:
ThreadEnv varPurpose
MainANTHROPIC_MODELPrimary codegen, conversation, complex tasks
HaikuANTHROPIC_DEFAULT_HAIKU_MODELFile watching, linter checks, completion, background jobs
Typical pairing: main thread on claude-opus-4-7 or claude-sonnet-4-6 for heavy lifting, haiku thread on claude-haiku-4-5-20251001 for background work — high quality at much lower token cost.
export ANTHROPIC_BASE_URL="https://utoken-test.uuzu.com"
export ANTHROPIC_AUTH_TOKEN="sk-zbTYx*******************************fceTvn5"
export ANTHROPIC_MODEL="claude-sonnet-4-6"
export ANTHROPIC_DEFAULT_HAIKU_MODEL="claude-haiku-4-5-20251001"
cd your-project-folder
claude

​
5.4 Single model (optional)

If you only want one model, there are two options: Option A — set both env vars to the same model 
export ANTHROPIC_BASE_URL="https://utoken-test.uuzu.com"
export ANTHROPIC_AUTH_TOKEN="sk-zbTYx*******************************fceTvn5"
export ANTHROPIC_MODEL="kimi-k2-250905"
export ANTHROPIC_DEFAULT_HAIKU_MODEL="kimi-k2-250905"
cd your-project-folder
claude
Option B — switch in-session Keep the dual-model config, start Claude Code, then in the interactive UI: 
/model haiku
The main thread switches to the model from ANTHROPIC_DEFAULT_HAIKU_MODEL.
ScenarioMain (ANTHROPIC_MODEL)Haiku (ANTHROPIC_DEFAULT_HAIKU_MODEL)
Production (top)claude-opus-4-7claude-haiku-4-5-20251001
Daily dev (recommended)claude-sonnet-4-6claude-haiku-4-5-20251001
Cost-optimizedkimi-k2-250905claude-haiku-4-5-20251001
Long contextclaude-sonnet-4-5-20250929 (200K)claude-haiku-4-5-20251001
Quick smoke-testclaude-haiku-4-5-20251001claude-haiku-4-5-20251001
See full IDs in the Utoken model catalog.

​
7. FAQ

  1. Check ANTHROPIC_AUTH_TOKEN for stray whitespace or newlines.
  2. Confirm the key is active in the Utoken console.
  3. Ensure your balance is positive.
For Claude Code, do not append /v1 to ANTHROPIC_BASE_URL. Use https://utoken-test.uuzu.com (Anthropic protocol differs from OpenAI).
In the interactive UI:
/model              # show current model
/model haiku        # switch to ANTHROPIC_DEFAULT_HAIKU_MODEL
/model sonnet       # alias supported in some versions
Or exit, change env vars, and re-launch.
Use the full ID from the Utoken model list, e.g.:
  • ✅ claude-sonnet-4-5-20250929
  • ❌ claude-sonnet-4-5
The version suffix is required.
  1. Prefer the dual-model setup with Haiku as the background thread.
  2. Use --max-turns and similar flags to cap conversation length.
  3. Maintain project memory under .claude/ to avoid repeating context.
  4. Use /clear to drop stale context.
Put env vars into a project .env, your shell profile, or direnv. Each engineer only swaps in their own ANTHROPIC_AUTH_TOKEN.

​
8. References