Agentic AI

AgenticAI.txt

#Codex
#Run Codex
```
codex
```
#To change model
```
/model
```

## What’s included in an AGENTS.md?
Common sections:
* Project overview and structure
* Build and test commands
* Helpful CLI tools and MCP servers
* Workflow for implementing a feature
* Pointers to other task-specific guidance for the agent

#Status | /status (Modle, Context Window, Used)
│ Visit https://chatgpt.com/codex/settings/usage for up-to-date                   │
│  Model:                gpt-5.4 (reasoning none, summaries auto)                 │
│  Directory:            E:\reactJS\resources-calendar                            │
│  Permissions:          Custom (workspace-write, on-request)                     │
│  Agents.md:            AGENTS.md                                                │
│  Account:              subroto.doict@gmail.com (Go)                             │
│  Collaboration mode:   Default                                                  │
│  Session:              019da3f3-7c32-7441-a637-450dd05cd23a                     │
│  Context window:       97% left (19.6K used / 258K)                             │
│  Weekly limit:         [███████████████████░] 97% left (resets 10:15 on 26 Apr) │

#agents.md create 1st time
- /init // first time you can use it. It use token, but not that good in analyzing project

# Generate/Regenerate
```
Scan my entire repository and regenerate AGENTS.md based on current codebase.

Include:
- Project structure (Next.js App Router)
- Key folders (app, components, lib, hooks)
- API patterns
- Coding conventions (TypeScript, naming)
- Build & run commands
- Any recent changes reflected in latest commits
Overwrite outdated sections.
```

#Gimini CLI

#Context- is the key, It remember session, GEMINI.md called context file. Without context, Gemini is just a smart stranger
- Suppose a drank intern come and he does now know how much sugar I take for a cup of tea. Context helps here
- It includes: Projects using javascript, Project structure, Rule to apply etc.

/memory show // show memory status
/init // to initialize the projects: It can see project, read packeg.json, to through some key files like index, different data, detalising the project structure and technologies are used

#Context Window and Context Files:
Context Window: Physical limit of how much the AI can "remember" at once (e.g., 1 million tokens for Gemini 1.5 Pro).
Context Files (gemini.md): This is a manual way to force specific information into that memory so the AI doesn't forget your project rules or tech stack

#Context and Model context: 
Think of Context as the "Brain's Memory" | Model Context as the "Nerve System" connecting the brain to the outside world.
Context: You put in your gemini.md (Static Knowledge), textbook you give the AI to read before the exam.
Model Context (MCP): A Standard System that connect Gemini directly to your local files, databases, or web tools automatically.
Analogy: Giving the AI a high-speed internet connection and a login to your computer so it can look things up itself.

#Gemini CLI Features
- Reads active cursor position, any text you have highlighted (up to 16KB), and 10 most recently accessed files
- Legacy codebases, the CLI's massive context window (up to 1 million tokens)
- Tackling massive system upgrades—such as migrating a 1,000-file legacy monolith from PHP 5 to PHP 8—the CLI can systematically parse directories, find deprecated syntax (like old mysql_ functions), and rewrite them to modern standards (like mysqli or PDO).
- Built-in web search and fetch tools. If you run into an obscure API error or need the latest documentation for a newly released library, the CLI can search the web and apply that real-time knowledge directly to your code.
- Model Context Protocol (MCP), allowing you to connect custom local servers, databases, and third-party tools to extend its capabilities even further.












```
```

```
```