Skip to content

Deprecated — Consolidated into DEV-0001 on 2026-05-02 per ADR-0047. This source file is retained as a reference; the canonical content is in DEV-0001.

ADR-0021: Claude Code as platform developer AI tooling

Status

Accepted — 2026-04-02

Supersedes: Partially overlaps with ADR-0011. ADR-0011 covers IaC authoring scope; this ADR covers developer tooling boundary with Ollama.

Context

The Archon platform spans Ansible, Terraform, k3s, Azure DevOps pipelines, OT firmware, and documentation. Development occurs on the CanEast AI Node (WSL) using a single-operator model. The volume and breadth of work across infrastructure, security, and automation warrants AI-assisted development tooling.

The platform already runs a local AI stack (Ollama, Qwen3:4b) suitable for general queries, but local models at this parameter scale lack sufficient context retention and reasoning depth for complex, multi-file infrastructure tasks — such as auditing an entire Ansible role for Ubuntu 25.10 compatibility, tracing lint violations across a role tree, or authoring ADRs that reference current repo state.

Decision

Claude Code (Anthropic CLI) is the AI developer assistant for the Archon platform.

  • Installed on: CanEast AI Node — WSL
  • Working directory: ~/homelab/repos/
  • Model: Claude Sonnet 4.6 (claude-sonnet-4-6)
  • Scope: Ansible role authoring and auditing, ADR writing, PR creation, Azure DevOps CLI operations, ansible-lint triage, general IaC work

Claude Code operates interactively in the terminal with access to the full repo context, git history, and shell tools. It does not replace the local Ollama stack — it complements it for tasks requiring deeper reasoning or multi-file context.

Boundary with local AI stack

Use case Tool
Quick lookups, chat, exploratory questions Ollama / Open WebUI (Qwen3:4b)
Ansible role authoring, auditing, lint fixes Claude Code
ADR writing, PR descriptions, doc updates Claude Code
Azure DevOps CLI operations from terminal Claude Code
OT firmware (ESP32, Arduino) Claude Code or Ollama (TBD)

Alternatives Considered

Ollama (Qwen3:4b) for all tasks — Already running locally. Sufficient for chat and quick lookups. Insufficient context window and reasoning depth for multi-file infrastructure audits and complex ADR authoring at this stage.

GitHub Copilot — IDE-only, no terminal/CLI integration, weak for infrastructure-as-code beyond code completion. Does not support interactive multi-step tasks.

No AI tooling — Single-operator platform with broad scope. AI assistance materially reduces time-to-deploy for non-trivial tasks like Ubuntu compatibility audits.

Consequences

  • Anthropic API usage billed to the account associated with Claude Code
  • PATs and sensitive credentials must not be typed into the Claude Code prompt — use env vars and dedicated credential helpers
  • Claude Code session context is not persistent across restarts — the ~/homelab/repos/archon-platform/.claude/ memory directory is used for cross-session context
  • Claude Code does not have access to the internal network — all Ansible runs and Azure DevOps CLI calls are executed locally and results are passed back in-session

References

  • https://claude.ai/code
  • CanEast AI Node: REDACTED, WSL, ~/homelab/repos/
  • ADR-0002 (Infisical) — credential handling relevant to AI tooling boundaries