Skip to content

16. OpenCode as Primary AI-Assisted Development Environment

Date: 2026-02-19

Status

Draft (Revised)

Context

SMUNZ and OpsDev.NZ development work involves: - Python CLI tool development (yalh-opsdevnz, specmaint-opsdevnz, etc.) - Infrastructure automation and configuration - Documentation and specification writing - Limited human capacity (single developer with AI assistance)

We need an AI-assisted development environment that: - Supports multiple AI model providers - Allows cost-effective model selection - Integrates well with our Linux-based development workflow - Can be extended for our specific needs

Decision

We will use OpenCode IDE as our primary AI-assisted development environment.

What is OpenCode?

OpenCode is an open-source IDE that provides: - AI assistant integration via the Zen service - Access to multiple model providers (free and paid) - Terminal integration for development commands - Extensibility for custom workflows

Why OpenCode?

Factor OpenCode Advantage
Cost flexibility Free tier models available; pay only when needed
Model variety Multiple providers through Zen service
Open source Can audit, extend, and contribute
Linux support Works well on our Fedora/AlmaLinux workstations
Simplicity Single environment for AI-assisted coding

Model Selection

We use models based on task requirements, not fixed assignments:

  • Routine development: Free models through Zen (current example: GLM variants)
  • Sensitive work: Privacy-focused paid models (Anthropic, etc.)
  • Sponsor collaboration: Sponsor-preferred platforms

Organisational AI usage policies govern model selection criteria for sensitive work.

Alternatives Considered

Alternative Why Not Chosen
Continue with ChatGPT Pro Personal account; organizational work should use organizational tools
Cursor IDE Proprietary; less flexibility in model selection
GitHub Copilot Subscription cost; limited model options
VS Code + extensions More configuration overhead; multiple extension ecosystems
Command-line only Less integrated experience; higher friction for routine work

Consequences

Positive

  • Single development environment for AI-assisted work
  • Cost flexibility: free models for routine work, paid when needed
  • Open source: can contribute improvements back
  • Good Linux support for our workstation environment
  • Simplified toolchain (one IDE for AI-assisted development)

Negative

  • Newer tool with smaller ecosystem than VS Code
  • Free model availability may change over time
  • Learning curve for team members unfamiliar with the tool

Mitigations

  • Document workflows and best practices
  • Maintain fallback options (VS Code, command-line tools)
  • Monitor free tier availability and adjust as needed

Implementation

  1. OpenCode installed on development workstations
  2. Zen service configured with available free models
  3. API keys managed via 1Password + op_opsdevnz tool
  4. Team documentation updated with OpenCode workflows
  5. Annual review of platform effectiveness
  • ADR #15: Evaluate AI Coding Platform for Development Productivity
  • Organisational AI usage policies govern tool selection and model choice