Blog
- How to plan with AI coding agents before they write code
Get a short plan from your coding agent before it edits code: what a good plan has, and three ways to do it, from plan mode to a plan step you approve.
- How to run multiple AI coding agents in parallel using Git worktrees
Run several coding agents at once without them overwriting each other: one git worktree per task, a small script, or a board that sets them up for you.
- How to stop your AI coding agent repeating the same mistakes
Put the correction in a rules file, a failing check, the task, or a skill. In Coppice, approve a Knowledge note so a later matching ticket lists the note.
- How to do multi-agent coding without losing control
Run more than one coding agent without losing the plan or the diff: terminals and worktrees, subagents in one CLI, or a desktop board you approve.
- How to use Fable for planning and a cheaper model for coding
Use Fable 5.2 for the plan and Opus or Sonnet for the code. Claude Code has no Fable-then-cheaper setting. Coppice can set one model per agent.
- How to use Claude Code and Codex together on one project
Three ways to run Claude Code and Codex on the same codebase: two terminals, git worktrees, or one board where each agent takes its own tickets.
- How to review AI-generated code before merging
What to check in an AI agent's diff before you merge it, and three ways to review agent code: in git, with a second agent, or with a plan you approve first.