DeeplyAgentic

Architecture, prompts, and tools for AI agent systems that actually work in production. 🚀 Harness AI. Reclaim Time. Amplify Impact.

🌐 Website · 💼 LinkedIn · ⚡ Quickstart

---

🔥 Trending This Week

| Project | Why it matters | |---------|---------------| | SkillSpector | NVIDIA's security scanner for agent skills — found malicious intent in 5% of bundles. Scan before you install. | | Skills Best Practices | The definitive guide to structuring agent skills — folder conventions, lazy loading, description triggers | | AI Engineering From Scratch | 230+ lessons across 20 phases — from linear algebra to autonomous agent swarms. Free. | | Claude Code Talks | 15 sessions from Anthropic's Code w/ Claude 2026 — subagents, MCP, scaling patterns | | Odysseus | Self-hosted AI workspace — chat, agents, deep research, email, calendar, memory — 65k ⭐ | | Deep Agents | LangChain's open-source Claude Code alternative — any LLM, $0, sub-agents, memory, checkpointing via LangGraph | | Mirage | Virtual filesystem that mounts S3, Drive, Slack, Gmail, GitHub, Notion, Postgres under one root — agents use bash, zero API translation | | CodeGraph | MCP server that cuts Claude Code tool calls by 92% — pre-indexes your codebase into a knowledge graph |

---

Get Started

# Option A: Local (no API keys needed)
git clone https://github.com/Mehtabk/DeeplyAgentic.git
cd DeeplyAgentic/quickstart
pip install ollama && python multi_agent_pipeline_ollama.py

# Option B: Cloud (OpenAI)
export OPENAI_API_KEY=your-key
uv run multi_agent_pipeline.py

---

Projects

| Project | Description | |---------|-------------| | [⚡ Quickstart](./quickstart/) | Run a multi-agent pipeline (Planner → Executor → Validator) in 5 minutes. | | [📚 The Agentic Stack](./the-agentic-stack/) | Curated tools for production agent systems — organized by architecture layer. | | [🧠 Agent Prompts](./agent-prompts/) | Production-ready system prompts — Planner, Executor, Validator, Orchestrator, Summarizer. | | [🔍 Comparisons](./comparisons/) | Side-by-side framework comparison: CrewAI vs AutoGen vs LangGraph vs OpenAI SDK vs Mastra. | | [✅ Checklists](./checklists/) | Pre-launch checklist for agent systems. Don't ship without it. | | [📐 Diagrams](./diagrams/) | Copy-paste Mermaid diagrams for agent architectures. | | [📝 Agent Decisions](./agent-decisions/) | Architecture Decision Records (ADRs) for agent systems. | | [🔁 Patterns](./patterns/) | Reusable architectural patterns — cost routing, the 8 canonical agent architectures, decision matrix. | | [📖 Reading List](./reading-list/) | Curated papers, reports, talks, and courses. Updated weekly. |

---

The 4-Layer Architecture

> Most agent projects fail not because of bad models — but because of missing architecture.

flowchart TB
    subgraph Orchestration["🎯 Orchestration Layer"]
        Router[Router] --> Coordinator[Coordinator]
    end
    subgraph Agents["🧠 Agent Layer"]
        A1[Planner] --> A2[Executor]
        A2 --> A3[Validator]
    end
    subgraph Tools["🔌 Tool Layer - MCP"]
        T1[APIs] --> T2[Databases]
        T2 --> T3[Services]
    end
    subgraph Guardrails["🛡️ Guardrails"]
        G1[Validation] --> G2[Escalation]
    end
    Orchestration --> Agents
    Agents --> Tools
    Guardrails -.->|enforces| Orchestration
    Guardrails -.->|enforces| Agents
    Guardrails -.->|enforces| Tools

---

Key Stats (May 2026)

| Stat | Source | |------|--------| | 80% of enterprise apps embed at least one AI agent | Gartner Q1 2026 | | 40% of agentic AI projects cancelled without architecture | Gartner | | 1,600+ agents per enterprise by year-end | IBM 2026 | | 1,445% surge in multi-agent inquiries | Gartner |

---

Community

| | | |---|---| | 💼 LinkedIn | Architecture breakdowns, tool reviews, and weekly agent ecosystem updates | | 🌐 Website | Landing page and resource hub | | 📰 [Changelog](./CHANGELOG.md) | Weekly updates to the repo |

> Want to be featured? If you're using DeeplyAgentic patterns in production, open a PR to add your story.

---

Contributing

We welcome contributions! See individual project folders for guidelines.

Quick ways to contribute:

• 🐛 Found a broken link? Open an issue
• 📚 Know a tool that belongs here? Submit a PR
• ⭐ Star this repo to help others find it

---

MIT License © 2026 DeeplyAgentic

Source & license

This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Mehtabk
• Source: Mehtabk/DeeplyAgentic
• License: MIT

Install and usage instructions live in the source repository linked above.