Z.AI Coding Plan Setup

Configure OpenClaw to use Z.AI GLM Coding Plan models with optimal settings, region selection, API key setup, model choice, and MCP tools for enhanced coding...

Security Scan

The instructions largely match a Z.AI integration for OpenClaw, but there are inconsistencies and missing declarations (notably an API key requirement) and the guide asks you to enable powerful agent tools (filesystem/shell/browser) without documenting risks — proceed only if you trust the source and understand the changes.

Purpose & Capability

The skill's name and description (configure OpenClaw to use Z.AI GLM models) match the SKILL.md content: it tells you how to set the endpoint, model, and config. However the registry metadata declares no required environment variables or primary credential while the instructions explicitly require a ZAI_API_KEY and endpoints — a clear metadata/instruction mismatch.

Instruction Scope

The SKILL.md tells the agent/user to edit ~/.openclaw/openclaw.json, set ZAI_API_KEY, enable tooling (filesystem, shell, browser), change config values (context tokens), restart the gateway, and 'install Z.AI MCP Tools'. Editing the user OpenClaw config and enabling shell/filesystem tools is within the functional scope of creating a coding agent but also grants broad local access. The guide does not show how MCP tools are installed and gives no verification steps for downloaded components.

Install Mechanism

This is instruction-only with no install spec or code files, so nothing will be automatically downloaded or written by the skill itself. That lowers installation risk. However the doc references installing MCP tools but provides no install commands or verified source URLs.

Credentials

The instructions require a ZAI_API_KEY environment variable and suggest endpoints for international/China regions, but the skill metadata declares no required env vars or primary credential. Requiring an API key is proportionate to the stated purpose, but the omission from metadata is an inconsistency that prevents automated gate checks and user warnings. Also enabling tools like 'filesystem' and 'shell' effectively escalates what the agent can access relative to simply using a model API key.

Persistence & Privilege

The skill does not request always:true and is user-invocable only. It instructs editing the user's OpenClaw config file (~/.openclaw/openclaw.json) and restarting the gateway, which is normal for configuring a runtime integration and is limited to the user's OpenClaw instance rather than system-wide privilege escalation.

What to consider before installing

This skill's instructions look like a legitimate guide to configure OpenClaw with Z.AI models, but several things don't add up and you should be cautious: (1) The guide requires a ZAI_API_KEY but the skill metadata doesn't declare any required credentials — ask the publisher to declare the env var so automated checks can warn you. (2) The instructions ask you to enable tools like filesystem and shell for the agent; those grant the agent broad local access — only enable them if you trust the model provider and the agent's behavior. (3) The SKILL.md mentions installing MCP tools but includes no verified install commands or URLs; before running any install steps, verify official installation instructions from Z.AI or bigmodel.cn and prefer official release pages. (4) Back up ~/.openclaw/openclaw.json before making edits and validate configs with openclaw config validate. If you do not trust the skill source or cannot verify the endpoints and MCP tools, do not apply the changes. If you want to proceed safely, request the author to (a) add ZAI_API_KEY to the skill's declared requirements, (b) provide explicit, verifiable install steps or links for MCP tools, and (c) document why shell/filesystem/browser tools are needed and how to limit their scope.

Overview

Configure OpenClaw to use Z.AI GLM Coding Plan models for optimal coding performance. Z.AI provides multiple GLM models optimized for development tasks with varying context windows and performance characteristics.

Prerequisites

Before configuring Z.AI Coding Plan:

1. OpenClaw installed - Verify with openclaw gateway status
2. Z.AI API Key - Available from https://z.ai/subscribe (international) or https://bigmodel.cn (China)
3. Docker environment - API key available as ZAI_API_KEY environment variable

Available Models

Model ID	Context	Best For
glm-5	205K tokens	Latest flagship, best overall performance
glm-5-turbo	200K tokens	Fast responses, quick iterations
glm-4.7	205K tokens	Balanced performance, strong reasoning
glm-4.7-flash	200K tokens	Quick iterations, prototyping
glm-4.7-flashx	200K tokens	Optimized speed
glm-4.6	205K tokens	Stable, excellent coding
glm-4.6v	128K tokens	Vision + coding (UI work)
glm-4.5	131K tokens	Efficient coding
glm-4.5-air	131K tokens	Lightweight operations
glm-4.5-flash	131K tokens	Fast coding, cost-effective

Configuration Steps

Step 1: Verify OpenClaw Status

openclaw gateway status

Step 2: Choose Configuration Method

Option A: Interactive Setup (Recommended)

International users:

openclaw onboard --auth-choice zai-coding-global

China region users:

openclaw onboard --auth-choice zai-coding-cn

This prompts for the API key and configures everything automatically.

Option B: Manual Configuration

Edit ~/.openclaw/openclaw.json:

{
  "env": {
    "ZAI_API_KEY": "sk-your-api-key-here"
  },
  "agents": {
    "defaults": {
      "model": {
        "primary": "zai/glm-5"
      }
    }
  }
}

Important: Model format is zai/<model-name>.

Step 3: Configure Coding-Specific Settings

For optimal coding performance:

# Enable tool streaming (better for coding workflows)
openclaw config set agents.defaults.models."zai/glm-5".params.tool_stream true --json

# Set higher context for complex projects (optional)
openclaw config set agents.defaults.maxContextTokens 150000

Step 4: Optional - Create Coding-Focused Agent

For dedicated coding agents:

openclaw agents create coding-assistant

Configure in ~/.openclaw/openclaw.json:

{
  "agents": {
    "coding-assistant": {
      "model": {
        "primary": "zai/glm-5"
      },
      "systemPrompt": "You are an expert software developer. Focus on writing clean, efficient, and well-documented code. Always explain your reasoning and provide best practices.",
      "tools": ["filesystem", "shell", "browser"]
    }
  }
}

Step 5: Restart and Verify

# Validate configuration
openclaw config validate

# Restart gateway
openclaw gateway restart

# Check status
openclaw status

Step 6: Test Configuration

Test with a simple coding task:

openclaw chat "Write a Python function to calculate the fibonacci sequence"

Model Selection Guide

Use the right model based on your needs:

• General coding → zai/glm-5 - Best overall performance
• Quick prototyping → zai/glm-5-turbo - Fast responses
• Complex refactoring → zai/glm-4.7 - Strong reasoning
• Code review → zai/glm-4.6 - Stable, thorough
• UI/Frontend work → zai/glm-4.6v - Vision capabilities
• Simple scripts → zai/glm-4.5-flash - Cost-effective

Region Configuration

• International: zai-coding-global endpoint → https://api.z.ai/api/coding/paas/v4
• China: zai-coding-cn endpoint → https://api.bigmodel.cn/api/coding/paas/v4

Troubleshooting

Authentication Error (1001)

API key missing or invalid:

openclaw config get env.ZAI_API_KEY

Model Not Found

Ensure correct endpoint:

# International
https://api.z.ai/api/coding/paas/v4

# China
https://api.bigmodel.cn/api/coding/paas/v4

Rate Limiting

GLM Coding Plan has usage limits based on subscription tier:

• Lite: 120 prompts per 5-hour cycle
• Pro: 600 prompts per 5-hour cycle
• Max: Higher limits

Gateway Won't Start

# Check logs
openclaw gateway logs

# Validate config
openclaw config validate

# Reset if needed
openclaw config reset

Switching Models

To change the default model:

# Set new model
openclaw config set agents.defaults.model.primary "zai/glm-4.7"

# Restart to apply
openclaw gateway restart

Integration with Channels

This configuration works with all OpenClaw channels:

• Feishu: Coding model used for all coding requests
• DingTalk: Same model configuration
• WeCom: Compatible with coding workflows
• CLI: Direct access via openclaw chat

Useful Commands

Command	Description
openclaw config get agents.defaults.model	Check current model
openclaw config set agents.defaults.model.primary "zai/glm-5"	Change model
openclaw gateway restart	Apply changes
openclaw status	Verify configuration

Z.AI MCP Tools Configuration

GLM Coding Plan includes powerful MCP tools for web access and code research. These tools are exclusive to Coding Plan subscribers.

Available MCP Tools

Tool	Description	Use Case
Web Search	Search the web for real-time information	Finding documentation, news, tutorials
Web Reader	Extract full content from any webpage	Reading articles, docs, extracting data
Zread	Access GitHub repositories documentation & code	Understanding open-source projects

MCP Quotas by Plan

Plan	Web Searches + Readers + Zread
Lite	100 total
Pro	1,000 total
Max	4,000 total

---

Step 7: Install Z.AI MCP Tools

After configuring the model, add the MCP tools to enhance your coding agent.

Tool 1: Web Search (web-search-prime)

Provides real-time web search capabilities.

One-click install (OpenClaw):

openclaw mcp add web-search-prime --type http --url "https://api.z.ai/api/mcp/web_search_prime/mcp" --header "Authorization: Bearer YOUR_API_KEY"

Manual configuration - Add to ~/.openclaw/openclaw.json:

{
 "mcpServers": {
 "web-search-prime": {
 "type": "http",
 "url": "https://api.z.ai/api/mcp/web_search_prime/mcp",
 "headers": {
 "Authorization": "Bearer YOUR_API_KEY"
 }
 }
 }
}

Available tool: webSearchPrime - Search web information with titles, URLs, summaries

---

Tool 2: Web Reader (web-reader)

Extracts complete content from any webpage.

One-click install (OpenClaw):

openclaw mcp add web-reader --type http --url "https://api.z.ai/api/mcp/web_reader/mcp" --header "Authorization: Bearer YOUR_API_KEY"

Manual configuration - Add to ~/.openclaw/openclaw.json:

{
 "mcpServers": {
 "web-reader": {
 "type": "http",
 "url": "https://api.z.ai/api/mcp/web_reader/mcp",
 "headers": {
 "Authorization": "Bearer YOUR_API_KEY"
 }
 }
 }
}

Available tool: webReader - Fetch page title, main content, metadata, links

---

Tool 3: Zread (zread)

Access GitHub repositories for documentation and code analysis.

One-click install (OpenClaw):

openclaw mcp add zread --type http --url "https://api.z.ai/api/mcp/zread/mcp" --header "Authorization: Bearer YOUR_API_KEY"

Manual configuration - Add to ~/.openclaw/openclaw.json:

{
 "mcpServers": {
 "zread": {
 "type": "http",
 "url": "https://api.z.ai/api/mcp/zread/mcp",
 "headers": {
 "Authorization": "Bearer YOUR_API_KEY"
 }
 }
 }
}

Available tools:

Tool	Description
search_doc	Search documentation in GitHub repositories
get_repo_structure	Get directory structure and file list
read_file	Read complete code content of files

---

Complete MCP Configuration Example

Here's a complete ~/.openclaw/openclaw.json with all three MCP tools:

{
 "env": {
 "ZAI_API_KEY": "sk-your-api-key-here"
 },
 "agents": {
 "defaults": {
 "model": {
 "primary": "zai/glm-5"
 }
 }
 },
 "mcpServers": {
 "web-search-prime": {
 "type": "http",
 "url": "https://api.z.ai/api/mcp/web_search_prime/mcp",
 "headers": {
 "Authorization": "Bearer sk-your-api-key-here"
 }
 },
 "web-reader": {
 "type": "http",
 "url": "https://api.z.ai/api/mcp/web_reader/mcp",
 "headers": {
 "Authorization": "Bearer sk-your-api-key-here"
 }
 },
 "zread": {
 "type": "http",
 "url": "https://api.z.ai/api/mcp/zread/mcp",
 "headers": {
 "Authorization": "Bearer sk-your-api-key-here"
 }
 }
 }
}

---

Step 8: Verify MCP Tools Installation

After adding the MCP tools, restart and verify:

# Restart gateway to load MCP servers
openclaw gateway restart

# Check MCP status
openclaw mcp list

# Verify tools are available
openclaw status

---

Step 9: Test the MCP Tools

Test each tool to ensure proper configuration:

Test Web Search:

openclaw chat "Search for the latest Python 3.12 features"

Test Web Reader:

openclaw chat "Read and summarize the content from https://docs.python.org/3/whatsnew/3.12.html"

Test Zread:

openclaw chat "Explain the structure of the fastapi/fastapi repository on GitHub"

---

MCP Tools Troubleshooting

Invalid API Key

Error: Invalid access token

Solutions:

1. Verify API key is correctly copied (no extra spaces)
2. Check API key is activated at https://z.ai/manage-apikey/apikey-list
3. Ensure sufficient balance on your account
4. Verify Authorization: Bearer format is correct

Connection Timeout

Solutions:

1. Check network connection
2. Verify firewall allows HTTPS connections
3. Increase client timeout settings
4. Try again later (server may be busy)

Repository Access Failed (Zread)

Solutions:

1. Confirm repository exists and is public
2. Check repository name format: owner/repo
3. Visit https://zread.ai to verify repository is indexed

Quota Exceeded

Error: Rate limit exceeded

Solution: Wait for the next 5-hour cycle or upgrade your plan

---

MCP Tools Use Cases

Scenario	Tools to Use	Example Prompt
Research a library	Zread + Web Search	"Research the React hooks API and find best practices"
Debug an issue	Web Search + Web Reader	"Search for solutions to 'Python asyncio CancelledError'"
Learn a new framework	Zread	"Explain the architecture of the Next.js repository"
Stay updated	Web Search	"What are the latest updates in TypeScript 5.4?"
Read documentation	Web Reader	"Read the FastAPI authentication documentation"

References

• Z.AI Coding Plan: https://z.ai/subscribe
• OpenClaw Z.AI Provider: https://docs.openclaw.ai/providers/zai
• GLM Models Overview: https://docs.openclaw.ai/providers/glm
• Web Search MCP: https://docs.z.ai/devpack/mcp/search-mcp-server
• Web Reader MCP: https://docs.z.ai/devpack/mcp/reader-mcp-server
• Zread MCP: https://docs.z.ai/devpack/mcp/zread-mcp-server