Coding with AI agents

This group is about getting a coding agent to work in your own repository: choose the agent, work out what it costs, give the repository a contract, design code the agent can handle, and run several agents in parallel when one is not enough. Subscriptions and free allowances are in step 2; per-token prices are in the Model & API Radar, and models you can run yourself are in Local Models for Coding Agents.

Reading Order

Step	Article	Your job
1	Choosing an AI Coding Agent	Choose the agent: what public scores show, and how to compare agents on your own tasks.
2	Free Allowances for AI Coding CLIs	Work out the cost: genuinely free allowances, subscription usage, API billing and temporary promotions.
3	Repository-Centered Codex Workflow	Give the repository a contract: directories, AGENTS.md, architecture notes and tests tell the agent how to work.
4	Software Design for AI Agents	Design for agents: how ubiquitous language, TDD, deep modules and gray-box delegation cut rework.
5	Multi-Agent Coordination: Subagents, Worktrees, Reviewers, and Event Logs	Run agents in parallel, and decide who merges and who verifies.
6	Herdr: A Persistent Terminal for Several Coding Agents	Keep several coding-agent sessions in persistent terminals.
7	How to Learn with AI	When learning something new, let AI find gaps, explain one step and set practice, while recall and derivation stay with you.

My reading