Trae AGENTS, MCP, RULES & New Upgrades: This FULLY FREE AI Coder IS REALLY GOOD!

Trey Updates: Comprehensive Summary

Key Concepts:

• Trey: A free code generation tool.
• Agents: Customizable AI assistants with specific instructions and tool access.
• MCP (Model-as-Code Platform) Servers: Allow integration of external tools and functionalities.
• Context: Documents and web URLs used to provide information to the AI.
• Rules: User-defined constraints and guidelines for code generation (global and project-specific).
• Auto-run: Automatic execution of commands and MCP tools by agents.
• Deny List: A list of commands that agents are prohibited from executing.

1. Introduction and Free Tier Considerations

Trey is a free code generation tool that utilizes generated code for training purposes, similar to Gemini's free tier API, Bolt, and V0. While free, users should be aware of this data usage.

2. Expanded Model Availability

Trey has significantly expanded its model offerings. Previously offering Claude 3.7 Sonnet, it now includes:

• Gemini 1.5 Pro
• GPT-4.1
• Deepseek V3.1

These models are available for free without usage limits.

3. Unified Chat and Builder Panel

The separate chat and builder panels have been merged into a single, unified interface, streamlining the user experience.

4. Agent Customization and Functionality

• Concept: Agents are AI assistants similar to those in Cursor or Rue Code, but with enhanced customization.
• Customization: Users can define custom system prompts and instructions for each agent, tailoring their behavior for specific tasks (e.g., a designer agent for front-end design, a testing agent).
• Default Agent: A "builder" agent is provided by default, replicating the previous Trey functionality.
• MCP Server Configuration: Each agent can be configured to use specific MCP servers, enabling access to specialized tools and functionalities.

5. Contextual Input: Documents and Web URLs

• Documents: Users can upload documents (Markdown, text files) or provide URLs to documents. Up to 1,000 files are supported, with a total size limit of 50MB.
• Web URLs: Users can paste URLs, and the AI will automatically extract relevant content from the web pages to use as context.

6. Rules: Global and Project-Specific

• User Rules (Global): Rules based on personal preferences that apply to all projects.
• Project Rules: Rules specific to a single project, providing workspace-level control.
• Implementation: Rules are defined in files, allowing users to specify constraints and guidelines for code generation.

7. Auto-Run and Deny List

• Auto-Run: Agents can be configured to automatically execute commands and utilize MCP tools.
• Deny List: A list of commands that agents are prohibited from executing. Commands like "RM" are automatically included for safety.

8. MCP Marketplace

A new MCP marketplace, similar to Client's marketplace, allows users to easily add MCP servers to their agents.

9. Demonstration of New Features

The video demonstrates the new features in action, including:

• Model Selection: Showing the availability of Gemini 1.5 Pro and GPT-4.1.
• Agent Selection: Demonstrating how to switch between agents (builder, builder with MCP).
• MCP Server Addition: Showing how to add MCP servers through the marketplace or manually.
• Agent Creation: Walking through the agent creation interface, including naming, instruction setting, tool enabling/disabling, and MCP server assignment.
• Auto-Run Configuration: Showing how to enable auto-run and configure the deny list.
• Context Addition: Demonstrating how to add documents and web URLs as context.
• Rule Creation: Showing how to create global and project rules.

10. Custom Model Integration

Users can now add custom API keys for various providers and select any model, including through OpenRouter. Requests are routed through OpenRouter's servers, similar to Cursor's custom option.

11. Source Control Integration

The source control feature now includes the ability to generate commit messages based on the generated code.

12. Conclusion

The video concludes that the new updates to Trey are significant and improve its functionality. The speaker encourages viewers to share their thoughts and subscribe to the channel.

Notable Quotes:

• "Agents are similar to what you have in cursor or rue code but it's actually even better basically you can set a custom system prompt and instructions about how you want the agent to perform."
• "It's actually pretty amazing to see another thing is the option of rules previously this didn't exist in Trey there are two types of rules user rules and project rules."

Key Takeaways:

Trey has evolved into a more powerful and customizable code generation tool with the introduction of agents, improved context handling, rule-based constraints, and expanded model support. The free availability of advanced models like Gemini 1.5 Pro and GPT-4.1 makes it a compelling option for developers. The agent customization and MCP server integration provide a high degree of flexibility for tailoring the tool to specific tasks and workflows.