AI Agents & Automation MOC
AI Agents & Automation
Domain hub for AI coding agents, MCP integration, and automation workflows — covering tool selection, prompting methodology, and real-world deployment patterns.
All AI Agents
Tools & Concepts
| Type | Name | Topic |
|---|---|---|
| concept | Agentic Engineering Vocabulary & Patterns | AI Agents & Tools |
| concept | The Agentic Loop | AI Agents & Tools |
| concept | WhatsApp Automation | WhatsApp Automation |
| entity | Antigravity | Antigravity |
| entity | Claude AI | Claude AI |
| entity | Claude Code | AI Agents |
| entity | Codex | AI Agents |
| entity | Google AI Studio | AI Development |
| entity | NotebookLM | AI Agents |
| entity | VS Code Agent-First Development Series | AI Agents |
For a side-by-side feature comparison, see Master-Agent-Comparison-Matrix.
Topic Hubs
No results
Quick Decision Guide
| I want to... | Start here |
|---|---|
| Build a new feature or fix complex bugs | Claude Code |
| Run a 24/7 trading bot or WhatsApp agent | Automation |
| Connect AI to external tools (database, Gmail, trading platform) | MCP |
| Learn how to work with AI coding agents effectively | VS Code Agent-First Development Series |
| Compare all the major tools side-by-side | Master-Agent-Comparison-Matrix |
| Keep AI agents from breaking things | Safe AI Workflows - Mitigating Hallucinations and Overreach |
| Automate WhatsApp customer support or CRM | WhatsApp Automation |
| Understand Claude as a financial analysis platform | Claude AI |
MCP: The Universal Bridge
Model Context Protocol is the common thread across all these tools. It lets AI agents connect to external services — databases, email, trading platforms, WordPress — with read/write access. The vault covers MCP servers for Zerodha, TradingView, Gmail, WordPress, Obsidian, and custom DIY implementations.
Key distinction: community-built MCP servers often provide full read/write capabilities that official integrations restrict to read-only. This matters for automation where the agent needs to execute trades, send emails, or publish content.
Prompting & Methodology
The quality of agent output depends heavily on how you prompt. Core resources:
- claude-code-prompting-patterns — 10 patterns from Anthropic's official workshop (plan-first, feedback loops, CLAUDE.md, inline teaching)
- VS Code Agent-First Development Series — 5-part formula: harness, model, context, tools, prompts
- Safe AI Workflows - Mitigating Hallucinations and Overreach — the Audit-Verify-Plan protocol for safety-critical agent work
Common thread across all sources: start by understanding the codebase before editing, give the agent a feedback loop (tests, screenshots), and never let it operate without guardrails on production systems.
See Also
- Self-Hosting & Infrastructure MOC — the infrastructure these agents run on
- PKM & Obsidian MOC — the knowledge base agents read and write to
- Access Control Systems MOC — domain where AI agents manage hardware and access
- NotebookLM — Google's source-grounded research assistant, often paired with agents
- Obsidian — the vault that agents read and write to as shared memory
- N8N — workflow orchestration engine for connecting agents to APIs