ingpoc/claude-glm-agent

Claude GLM Agent Configuration

Configure any app to use z.ai GLM 4.7 proxy instead of standard Anthropic API.

Quick Start

# Initialize project for GLM configuration
~/.claude/skills/claude-glm-agent/scripts/init-glm.sh

# Configure .env files
./.claude/scripts/configure-glm.sh

# Test proxy connection
~/.claude/skills/claude-glm-agent/scripts/test-proxy.sh

# Migrate existing project from API_KEY to AUTH_TOKEN
./.claude/scripts/configure-glm.sh --migrate

Configuration Workflows

1. New Frontend Project (Next.js)

| Step | Action | |------|--------| | 1 | Detect package.json with Next.js or Vite | | 2 | Prompt for z.ai token | | 3 | Create .env.local with NEXT_PUBLIC_ prefix | | 4 | Add ANTHROPIC_BASE_URL pointing to z.ai proxy | | 5 | Add model overrides for GLM 4.7 | | 6 | Test proxy connection |

2. New Backend Project (Python/Node)

| Step | Action | |------|--------| | 1 | Detect requirements.txt or package.json | | 2 | Prompt for z.ai token | | 3 | Create .env without public prefix | | 4 | Add ANTHROPIC_BASE_URL and model overrides | | 5 | Test proxy connection |

3. Existing Project Migration

| Step | Action | |------|--------| | 1 | Scan for existing ANTHROPIC_API_KEY | | 2 | Backup current .env files | | 3 | Replace with ANTHROPIC_AUTH_TOKEN | | 4 | Add BASE_URL and model overrides | | 5 | Update code if needed (client factory pattern) | | 6 | Test and rollback on failure |

4. Netlify Deployment

| Step | Action | |------|--------| | 1 | Deploy via netlify-deploy skill | | 2 | Set environment variables via Netlify dashboard | | 3 | Configure ANTHROPIC_* variables for GLM 4.7 | | 4 | Redeploy to apply changes |

Environment Variables

Required Variables

| Variable | Value | Purpose | |----------|-------|---------| | ANTHROPIC_AUTH_TOKEN | Your z.ai token | Authentication with z.ai proxy | | ANTHROPIC_BASE_URL | https://api.z.ai/api/anthropic | Proxy endpoint |

Optional Model Overrides

| Variable | Value | Purpose | |----------|-------|---------| | ANTHROPIC_DEFAULT_HAIKU_MODEL | glm-4.7 | Route Haiku requests to GLM 4.7 | | ANTHROPIC_DEFAULT_SONNET_MODEL | glm-4.7 | Route Sonnet requests to GLM 4.7 | | ANTHROPIC_DEFAULT_OPUS_MODEL | glm-4.7 | Route Opus requests to GLM 4.7 |

Frontend vs Backend

| Environment | Variable Prefix | Example | |-------------|-----------------|---------| | Next.js (browser) | NEXT_PUBLIC_ | NEXT_PUBLIC_ANTHROPIC_AUTH_TOKEN | | Vite (browser) | VITE_ | VITE_ANTHROPIC_AUTH_TOKEN | | Backend (server) | None | ANTHROPIC_AUTH_TOKEN |

Configuration Examples

Next.js (.env.local)

# z.ai GLM 4.7 Proxy Configuration
NEXT_PUBLIC_ANTHROPIC_AUTH_TOKEN=your-zai-token-here
NEXT_PUBLIC_ANTHROPIC_BASE_URL=https://api.z.ai/api/anthropic
NEXT_PUBLIC_ANTHROPIC_DEFAULT_HAIKU_MODEL=glm-4.7
NEXT_PUBLIC_ANTHROPIC_DEFAULT_SONNET_MODEL=glm-4.7
NEXT_PUBLIC_ANTHROPIC_DEFAULT_OPUS_MODEL=glm-4.7

Vite (.env)

# z.ai GLM 4.7 Proxy Configuration
VITE_ANTHROPIC_AUTH_TOKEN=your-zai-token-here
VITE_ANTHROPIC_BASE_URL=https://api.z.ai/api/anthropic
VITE_ANTHROPIC_DEFAULT_HAIKU_MODEL=glm-4.7
VITE_ANTHROPIC_DEFAULT_SONNET_MODEL=glm-4.7
VITE_ANTHROPIC_DEFAULT_OPUS_MODEL=glm-4.7

Backend Python (.env)

# z.ai GLM 4.7 Proxy Configuration
ANTHROPIC_AUTH_TOKEN=your-zai-token-here
ANTHROPIC_BASE_URL=https://api.z.ai/api/anthropic
ANTHROPIC_DEFAULT_HAIKU_MODEL=glm-4.7
ANTHROPIC_DEFAULT_SONNET_MODEL=glm-4.7
ANTHROPIC_DEFAULT_OPUS_MODEL=glm-4.7

Backend Node.js (.env)

# z.ai GLM 4.7 Proxy Configuration
ANTHROPIC_AUTH_TOKEN=your-zai-token-here
ANTHROPIC_BASE_URL=https://api.z.ai/api/anthropic
ANTHROPIC_DEFAULT_SONNET_MODEL=glm-4.7
ANTHROPIC_DEFAULT_OPUS_MODEL=glm-4.7

Netlify Environment Variables

Set via dashboard or CLI:

netlify env:set ANTHROPIC_AUTH_TOKEN "your-zai-token"
netlify env:set ANTHROPIC_BASE_URL "https://api.z.ai/api/anthropic"
netlify env:set ANTHROPIC_DEFAULT_HAIKU_MODEL "glm-4.7"
netlify env:set ANTHROPIC_DEFAULT_SONNET_MODEL "glm-4.7"
netlify env:set ANTHROPIC_DEFAULT_OPUS_MODEL "glm-4.7"

Code Patterns

Client Factory Pattern

Best practice for environments requiring both standard and proxy API access:

// lib/anthropic-client.ts
import Anthropic from '@anthropic-ai/sdk'

const baseURL = process.env.ANTHROPIC_BASE_URL || 'https://api.anthropic.com'
const authToken = process.env.ANTHROPIC_AUTH_TOKEN || process.env.ANTHROPIC_API_KEY

export const anthropic = new Anthropic({
  baseURL,
  apiKey: authToken,
  defaultHeaders: {
    'anthropic-version': '2023-06-01'
  }
})

Python Client

# lib/anthropic_client.py
import os
from anthropic import Anthropic

base_url = os.getenv('ANTHROPIC_BASE_URL', 'https://api.anthropic.com')
auth_token = os.getenv('ANTHROPIC_AUTH_TOKEN') or os.getenv('ANTHROPIC_API_KEY')

anthropic = Anthropic(
    base_url=base_url,
    api_key=auth_token,
)

Troubleshooting

| Issue | Solution | |-------|----------| | 401 Unauthorized | Check ANTHROPIC_AUTH_TOKEN is valid z.ai token | | Connection timeout | Verify ANTHROPIC_BASE_URL is correct | | Model not found | Check model overrides, ensure GLM 4.7 is available | | CORS error (browser) | Ensure NEXT_PUBLIC_ or VITE_ prefix is used | | Proxy not working | Run test-proxy.sh to verify connection |

Migration Guide

From ANTHROPIC_API_KEY to AUTH_TOKEN

| Step | Action | |------|--------| | 1 | Replace ANTHROPIC_API_KEY with ANTHROPIC_AUTH_TOKEN | | 2 | Add ANTHROPIC_BASE_URL=https://api.z.ai/api/anthropic | | 3 | Add model override variables | | 4 | Update code to use AUTH_TOKEN (or use fallback pattern) | | 5 | Test with test-proxy.sh |

Fallback Pattern (Best Practice)

Support both API_KEY and AUTH_TOKEN:

const apiKey = process.env.ANTHROPIC_AUTH_TOKEN || process.env.ANTHROPIC_API_KEY

Scripts

| Script | Purpose | |--------|---------| | init-glm.sh | Initialize project: copy configure-glm.sh to .claude/scripts/ | | configure-glm.sh | Configure .env files (copied to project) | | test-proxy.sh | Test z.ai proxy connection | | migrate-env.sh | Migrate from API_KEY to AUTH_TOKEN |

Resources

| Resource | Load When | |----------|-----------| | env-patterns.md | Need environment variable patterns for specific frameworks | | client-config.md | Need client factory pattern examples | | frameworks.md | Need framework-specific setup instructions |