AI Coding Agent Orchestration

A reference architecture for agentic software development.

This configuration guide is an opinionated but practical handbook that argues for treating AI coding tools as a composable, multi-agent system rather than a single assistant, and then walks through how to implement that mindset using OpenCode's CLI, configuration system, and ecosystem (agents, tools, skills, and plugins). It emphasizes that productivity gains come from structured context, explicit orchestration, and layered configuration (global -> project -> runtime), enabling developers to build a repeatable “AI team” that plans, codes, reviews, and integrates work with minimal manual prompting.

If you want the shortest path through the material: read Part 1 for the mental model, follow Part 2 for the baseline setup, use Part 3 only for the integrations you actually need, and use Part 4 only after your setup has grown enough to justify orchestration cleanup.

Tip

Even if you already know the composition of an AI coding agent, Part 1 is still a good place to refresh that knowledge.

Note

This is an opinionated selection of tools and instructions based on personal experience. The AI tooling ecosystem changes quickly, so treat this as a starting point and adapt it to your own workflow.

MIT License

---

Who This Is For

• Developers who already use AI coding assistance in an IDE and want to understand when a CLI agent adds value.
• Engineers who want a repeatable OpenCode setup for local development, documentation, context, and external tools.
• Advanced users who want to extend coding agents with skills, MCP servers, LSPs, Jira, Azure DevOps, and Azure integrations.

Parts 1 and 2 are accessible to anyone interested in AI-assisted development. Part 3 is more advanced and focuses on optional tooling, custom skills, and integrations.

How To Read This Guide

Start with Part 1 if you are new to AI coding agents or if you want to refresh the concepts. Go directly to Part 2 if you already understand the basics and want the default setup. Use Part 3 as a reference when you are ready to extend the environment.

Suggested reading paths:

• New to coding-agent CLIs: Part 1 → Part 2
• Already using OpenCode and want a stronger setup: Part 2 → selected sections from Part 3
• Already installed several plugins, skills, and MCP servers: Part 4 after finishing Part 2 or Part 3

Guide Parts

1. Part 1: Why Use an AI Coding Agent CLI?
   Explains the evolution from chat and IDE completion to agentic CLI workflows, the agent loop, context management, and the tool landscape.

2. Part 2: A default setup
   Provides the practical setup path: installing OpenCode, adding context, configuring tools, installing skill packs, and adopting daily-use habits.

3. Part 3: Advanced Use
   Covers optional advanced additions such as engineering skills, custom review workflows, SQL LSP support, Jira workflows, Azure DevOps, and Azure MCP integration.

4. Part 4: Optimisation
   This part describes a couple of post install optimisations that you could do for eliminating overlap in skill packs and MCP servers, orchestration layers. There is also a section about optimising your model routing strategy and budget management.

Contribution

Questions, corrections, and suggestions are welcome. Open an issue at CodingAgentOrchestration issues or contribute with a pull request.

Useful contributions include:

• correcting stale install prompts or links
• adding verification steps for existing integrations
• documenting tradeoffs for tools that looked promising but did not hold up in practice
• improving cross-platform instructions where Windows, macOS, and Linux differ