GLM Code Plan Updates and Usage with Goose AI Agent
Key Concepts:
- GLM Code Plan: An alternative to Claude Code for AI coding and general AI tasks.
- MCP (Managed Compute Provider): Services like Vision MCP and Web Search MCP that enhance the capabilities of AI models.
- Goose AI: An open-source AI agent CLI tool used for design and general tasks.
- Kilo: An AI coding tool with a user-friendly interface.
- Recipes: Custom instructions for Goose AI agents to perform specific tasks.
- Shadcn/ui: A collection of re-usable UI components built using Radix UI and Tailwind CSS.
- Material UI: A popular React UI framework.
GLM Code Plan Overview
The video discusses the GLM Code plan as a cost-effective alternative to Claude Code, especially after Claude Code started offering quantized models even on their higher-priced plans. The GLM Code plan offers various tiers with different usage quotas and pricing structures.
- Pricing: Initially, plans were available at $3 and $15 per month, increasing to $6 and $30 in subsequent months. A new $30 plan is now available, increasing to $60.
- Max Plan: The presenter received free access to the Max plan from ZAI. This plan offers approximately 2,400 prompts every 5 hours, significantly more than Claude Code's Max 20X plan, at a lower cost.
- Quarterly and Yearly Plans: GLM Code now offers quarterly and yearly subscription options. The annual plan provides the initial discounted price for the entire year. For example, the $3 plan costs $36 annually instead of $72.
- Value Proposition: The $36 annual plan is highlighted as an excellent value, providing a substantial number of prompts (120 every 5 hours) compared to Claude Code Pro or ChatGPT Pro at a fraction of the cost.
Integration and Compatibility
The GLM Code plan is compatible with various AI coding tools, including:
- Goose
- Claude Code
- Open Code
- Rue Kilo Klein
The presenter uses Kilo for coding due to its user-friendly interface and ease of integration.
Managed Compute Providers (MCPs)
The Pro and Max plans include Vision MCP and Web Search MCP from ZAI at no extra cost.
- Vision MCP: Enables GLM to process images, create designs from inspiration, and perform OCR (Optical Character Recognition).
- Web Search MCP: Allows GLM to perform web searches without incurring additional costs.
Using GLM with Goose AI for Design
The presenter details how they integrated GLM Code with Goose AI to create a designer agent.
- Rationale: The presenter prefers using different tools for specific tasks and wanted to port their designer agent to GLM.
- Goose AI Selection: Goose AI was chosen over Open Code and Crush because it is more focused on being an AI agent, allowing for both chat interface and code generation capabilities.
- Setup: Goose AI offers both a desktop app and a CLI (Command Line Interface). The CLI is used for scripted workflows, while the desktop app is preferred for everyday use. Setting up GLM with Goose AI involves configuring settings as per the GLM documentation and enabling the Vision and Web Search MCPs.
- Recipes: Custom instructions (recipes) are created for Goose AI to define specific agent behaviors. The presenter created a designer agent that uses Shadcn/ui and Material UI for design tasks. A Context 7 MCP server is used to provide documentation and context.
Example Workflow
The presenter uses Goose AI to write code and create designs. The combination of GLM Code and Goose AI creates a personal AI assistant similar to Manis but local and open-source.
Key Arguments and Recommendations
- Value for Money: The GLM Code plan offers excellent value, especially the $36 annual plan.
- Flexibility: The GLM API can be configured in various coding tools.
- Goose AI Integration: Combining GLM Code with Goose AI creates a powerful and versatile AI agent.
- Recommendation: The presenter recommends the $36 annual plan for students and the $6 monthly plan for general users.
Conclusion
The GLM Code plan, particularly when combined with tools like Goose AI and the included MCPs, provides a compelling and cost-effective solution for AI-assisted coding and design. The presenter emphasizes the value proposition, flexibility, and ease of integration, making it a strong alternative to other AI coding platforms.
AI summaries can miss context or contain errors. Check important details against the original video.