# What Projects Are Included in the Capstone Projects Phase? A Complete Guide to Phase 19 > Discover the Capstone Projects phase (Phase 19) and build production-ready AI systems like coding agents and multi-agent pipelines. Integrate skills from prior phases for your portfolio. - Repository: [Rohit Ghumare/ai-engineering-from-scratch](https://github.com/rohitg00/ai-engineering-from-scratch) - Tags: how-to-guide - Published: 2026-08-30 --- **The Capstone Projects phase (Phase 19) includes portfolio-grade, production-ready AI systems such as terminal-native coding agents, end-to-end safety gates, and multi-agent orchestration pipelines that integrate skills from prerequisite phases.** The `rohitg00/ai-engineering-from-scratch` repository structures its curriculum into progressive phases, with Phase 19 serving as the culminating stage. This final phase moves beyond theoretical exercises to demand deployable, evidence-backed systems that demonstrate mastery of terminal-native agents, safety protocols, and autonomous workflows. ## Project Categories in the Capstone Projects Phase According to [`phases/19-capstone-projects/README.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/phases/19-capstone-projects/README.md), the curriculum groups capstone work into five distinct portfolio categories. While some projects like the Terminal-Native Coding Agent and End-to-End Safety Gate are fully implemented, others appear on the Phase 19 roadmap as future additions. ### Full-Stack Coding Agents The **Terminal-Native Coding Agent** represents the foundational capstone project. Located in [`phases/19-capstone-projects/01-terminal-native-coding-agent/code/main.py`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/phases/19-capstone-projects/01-terminal-native-coding-agent/code/main.py), this system demonstrates end-to-end agent orchestration capable of reading, writing, and executing code against real repositories. The project documentation in [`phases/19-capstone-projects/01-terminal-native-coding-agent/docs/en.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/phases/19-capstone-projects/01-terminal-native-coding-agent/docs/en.md) outlines learning objectives that require prior completion of Phases 11, 13, 14, 15, and 17. ### Safety-Oriented Pipelines The **End-to-End Safety Gate** project enforces budget constraints, event tracing, and model-level policies before production deployment. Implemented in [`phases/19-capstone-projects/87-end-to-end-safety-gate/code/main.py`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/phases/19-capstone-projects/87-end-to-end-safety-gate/code/main.py), this pipeline monitors trace events and policy violations as documented in [`phases/19-capstone-projects/87-end-to-end-safety-gate/docs/en.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/phases/19-capstone-projects/87-end-to-end-safety-gate/docs/en.md). The system serves as a concrete implementation of safety "gates" that prevent unauthorized model releases. ### Multi-Agent Orchestration Future roadmap items include complex systems that coordinate multiple specialized agents—such as planners, executors, and evaluators—to solve hierarchical tasks. These projects emphasize stitching disparate agents into cohesive workflows capable of handling complex, real-world scenarios. ### Autonomous Systems The curriculum targets deployable bots and services—including voice assistants and vision-driven robots—designed for continuous operation with minimal human intervention. These capstones focus on long-term stability and autonomous decision-making loops. ### Production-Grade Infrastructure Advanced capstones incorporate **Model-Context Protocol (MCP)** servers, comprehensive observability stacks, and automated deployment scripts. These projects create truly production-ready artifacts rather than disposable tutorial code. ## Universal Requirements for Capstone Completion Every project in Phase 19 adheres to three strict standards documented in the repository root and Phase 19 README: **Prerequisite-Driven Development** Learners must first complete specific earlier phases. Following [`phases/19-capstone-projects/README.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/phases/19-capstone-projects/README.md), the Terminal-Native Coding Agent explicitly requires knowledge from Phases 11, 13, 14, 15, and 17 before attempting implementation. **Evidence-First Documentation** Each submission must record specific telemetry: the command run, exit code, final plan state, budget totals, trace events, and at least one design decision that would change before connecting to a real model. This requirement ensures reproducible engineering practices and audit trails. **Artifact-Driven Deliverables** Unlike tutorial exercises, capstones must ship reusable components—whether prompts, skills, standalone agents, or MCP servers—that integrate into downstream production workflows. This aligns with the repository principle that every lesson ships something tangible. ## Running the Example Capstone Projects Execute the starter coding agent from the repository root: ```bash python3 phases/19-capstone-projects/01-terminal-native-coding-agent/code/main.py ``` Run the safety gate implementation: ```bash python3 phases/19-capstone-projects/87-end-to-end-safety-gate/code/main.py ``` Both commands generate comprehensive output including execution traces, budget consumption metrics, and policy violation reports, satisfying the evidence-first requirement specified in the Phase 19 documentation. ## Summary - The Capstone Projects phase (Phase 19) serves as the culminating stage of the `ai-engineering-from-scratch` curriculum, requiring integration of skills from prerequisite phases 11, 13, 14, 15, and 17. - Available implementations include the Terminal-Native Coding Agent (`01-terminal-native-coding-agent`) and End-to-End Safety Gate (`87-end-to-end-safety-gate`), with future additions covering multi-agent orchestration and autonomous systems. - All capstones demand evidence-based documentation, reusable artifact creation, and demonstrate production-grade architecture patterns. - Source implementations reside in `phases/19-capstone-projects/`, with specific logic separated into [`code/main.py`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/code/main.py) files and detailed documentation in [`docs/en.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/docs/en.md) subdirectories. ## Frequently Asked Questions ### What are the prerequisites for starting the Capstone Projects phase? According to [`phases/19-capstone-projects/README.md`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/phases/19-capstone-projects/README.md), learners must complete Phases 11, 13, 14, 15, and 17 before attempting the Terminal-Native Coding Agent or other capstone projects. These prerequisites ensure competency in terminal-native development, safety protocols, and agent orchestration fundamentals. ### How does the Evidence-First requirement work in practice? You must document the specific command executed, its exit code, the final plan state, total budget consumption, trace events, and one design modification you would implement before connecting to a real production model. This documentation standard applies to all projects within `phases/19-capstone-projects/` and ensures engineering rigor. ### What is the difference between the Terminal-Native Coding Agent and the End-to-End Safety Gate? The Terminal-Native Coding Agent focuses on autonomous code manipulation within repositories, implemented in [`phases/19-capstone-projects/01-terminal-native-coding-agent/code/main.py`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/phases/19-capstone-projects/01-terminal-native-coding-agent/code/main.py). The End-to-End Safety Gate emphasizes pre-deployment policy enforcement and budget monitoring, located in [`phases/19-capstone-projects/87-end-to-end-safety-gate/code/main.py`](https://github.com/rohitg00/ai-engineering-from-scratch/blob/main/phases/19-capstone-projects/87-end-to-end-safety-gate/code/main.py). ### Are the Capstone Projects suitable for production deployment? While designed as educational portfolio pieces, these projects follow production-grade standards including MCP server integration, comprehensive observability, and artifact-driven architecture. The End-to-End Safety Gate specifically demonstrates production-ready safety protocols, though additional hardening would be required for enterprise deployment.