ERP techstack =>


backend
golang + postgresql + gin web framework

frontend
react => =>typescript + tanstack query, router, table, caching=> fetch data, pindah page ngak load ulang

#stack #tech #backend #frontend #golang #project #web
https://developers.openai.com/codex/cli

https://agents.md/

https://jules.google/ => coding agents

https://modelcontextprotocol.io/docs/getting-started/intro => Model Context Protocol

https://youtu.be/hM06KRKteu8?si=e_YGNq_FQCUnuEbI => agents.md users

https://kilocode.ai/ => coding agent
https://kilocode.ai/pricing » tariff
https://youtu.be/pO7zRLQS-p0 » tutorial kilocode.ai


#codex #openai #chatgpt #agent #md #agents #coding #asisten #mcp #readme

A simple, open format for guiding coding agents,
used by over 20k open-source projects.

Think of AGENTS.md as a README for agents: a dedicated, predictable place to provide the context and instructions to help AI coding agents work on your project.
Here’s a breakdown of TRAE / Trae AI (sometimes called “Trae IDE” or “Trae Agent”) — its strengths, weaknesses, and how it compares with other AI-assisted coding tools. If you want a side-by-side comparison with a specific tool (e.g. GitHub Copilot, Cursor, etc.), I can do that too.

---

## What is TRAE / Trae AI / Trae Agent?

* TRAE (or Trae AI / Trae IDE) is an AI-powered development environment / coding assistant built (or backed) by ByteDance. ([trae.ai][1])
* It offers features such as:

* Builder / Agent mode: where you describe tasks or goals and the system plans and executes code changes. ([DataCamp][2])
* Chat / code assistance: inline suggestions, code completions, debugging help, etc. ([DataCamp][2])
* Model support: it gives access to strong LLMs like Claude 3.5 Sonnet, GPT-4o, etc. ([DataCamp][2])
* Multimodal / additional tools: image-to-code (uploading designs and converting to code), preview windows, built-in terminal, context indexing, etc. ([YouTube][3])
* Open agent / open source component: Trae Agent (the autonomous agent side) is open source. ([arXiv][4])
* The academic paper “Trae Agent: An LLM-based Agent for Software Engineering with Test-time Scaling” describes a modular agent architecture and shows strong benchmark performance. ([arXiv][4])

Because it’s relatively new, some features are still evolving, and access to certain “premium” capabilities or modes might be limited or gated. ([DataCamp][2])

---

## Pros (Strengths) of TRAE

Here are the key advantages that users and reviewers often point out:

| Strength | Explanation / Evidence |
| ------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------- |
| Free or very generous access (for now) | TRAE is currently offered at no cost (all features available) in many reports. ([Builder.io][5]) |
| Access to powerful models | Even in its free tier, it gives access to Claude 3.5, GPT-4o, etc. ([futuretools.io][6]) |
| Builder / “think-before-doing” approach | The Builder mode first plans and breaks down tasks before making code changes, which can reduce errors or misinterpretations. ([Builder.io][5]) |
| Rich context & indexing | It supports indexing existing code, project context, URL/document context to guide suggestions in a more informed way. ([DataCamp][2]) |
| Multimodal features | The ability to upload images (e.g. mockups) and convert them to UI code is a distinctive feature. ([YouTube][3]) |
| Benchmark & research backing | According to the “Trae Agent” paper, it outperformed certain baselines for software issue resolution on a repository-level benchmark. ([arXiv][4]) |
| Open agent / transparency | The agent component (Trae Agent) is open source, which helps with transparency, trust, and potential customization. ([arXiv][4]) |

Some user feedback also notes that the UI/UX and speed of prototyping are attractive compared to alternatives. ([Medium][7])

---

## Cons (Weaknesses / Risks / Limitations)

However, TRAE is not perfect. Here are some drawbacks and cautions: