9 AI Agent Design Patterns Based on Google ADK
MaryamMiradi · x · 2026-07-20
The author shares hands-on experience using Google ADK to build an end-to-end tax and accounting AI Agent system, summarizing 9 design patterns: **Single Agent Patterns (Layer 1)** - **Tool Use**: Parsing PDFs with Docling to extract invoice data - **ReAct Pattern**: Conducting tax code research through iterative reasoning loops - **Structured Output**: Generating financial reports using Pydantic schema - **Memory Augmentation**: Managing client accounts using session state **Composite Agent Patterns (Layer 2)** - **Reflection/Self-Critique**: Validating tax returns using `LoopAgent` - **Plan and Execute**: Preparing for audits using `SequentialAgent` - **Router/Dispatcher**: Routing tax queries via sub-agents - **Parallel Research**: Conducting multi-jurisdictional research using `ParallelAgent` - **Orchestrator-Worker**: Combining all ADK primitives to build a complete accounting pipeline
More from coding & agent
- AI makes software easier to build, but it also lowers the floor on quality — paw_lean · 2026-07-21
- Fable coding run costs $6.69 for 67 lines of code in a 4-minute job — bytebot · 2026-07-21
- Cursor writes better code, but ChatGPT can still control the computer — vista8 · 2026-07-21
- Agents can remember facts, but still forget how to do the job — No_Advertising2536 · 2026-07-21
- Agent skills for project downgrade and troubleshooting tested in CLAD on LS 5.22 — stspanho · 2026-07-21
- Open-source B-roll skill turns scripts into 5-second vertical clips with Codex and Gemini — yangyi · 2026-07-21