LangGraph Agent Skill

This skill provides a LangGraph-based agent framework for consistent tool calling across all OpenAI-compatible LLM providers.

When to Use

• You need guaranteed tool calling consistency (especially with GLM-5, Zhipu, or other models with inconsistent tool invocation)
• You want automatic tool loops that keep calling tools until the task is complete
• You're working with multiple LLM providers and need a unified interface
• You need to build custom agents with specific tool sets

Capabilities

• create_agent: Create a LangGraph agent with custom tools
• run_agent: Execute a task with the agent
• add_tool: Add a custom tool to an agent
• get_agent_result: Retrieve the final result from agent execution

Environment Variables

Required for the agent to function:

• LLM_API_KEY: Your LLM provider API key
• LLM_BASE_URL (optional): Custom base URL for OpenAI-compatible providers (default: OpenAI)
• LLM_MODEL (optional): Model name (default: "gpt-4o")

Provider Examples

# Z.AI / GLM-5
LLM_BASE_URL=https://api.z.ai/api/coding/paas/v4
LLM_MODEL=glm-5

# OpenRouter
LLM_BASE_URL=https://openrouter.ai/api/v1
LLM_MODEL=meta-llama/llama-3-70b-instruct

# Groq
LLM_BASE_URL=https://api.groq.com/openai/v1
LLM_MODEL=llama3-70b-8192

# Ollama (local)
LLM_BASE_URL=http://localhost:11434/v1
LLM_MODEL=llama3

# DeepSeek
LLM_BASE_URL=https://api.deepseek.com
LLM_MODEL=deepseek-chat

Usage Pattern

1. Create an agent with the tools you need
2. Run the agent with a task prompt
3. The agent automatically loops through tool calls until completion
4. Retrieve the final result