AutoKaam Playbook
Claude Code, the CLI Agent I Run All Day
Anthropic's terminal-first coding agent, and my daily driver for code work.
Last reviewed:
The operator take
Claude Code is my primary coding interface. On 28 Sep 2026, claude --version on my Ubuntu 24.04 box prints 2.1.283, installed with Anthropic's native installer, and it runs across every repo I ship from. It is the difference between a solo founder and something that feels like a team.
What I love: the terminal-first ergonomics. I do not want a separate IDE pane; I want my shell, my repo, my git status, and an agent that can read, edit, run, and commit in that same flow. Claude Code does this. It can hand parallel work to subagents, run shell commands behind permission gates, and ship clean diffs.
The 2.1 series shipped subagent forking, and that changed how I scope big migrations: I fan out parallel work streams from one orchestrator session, and the parent reads the summaries instead of every transcript.
What I have got wrong: trusting first-pass refactors without reading the diff. Claude Code is excellent but not infallible, and a rubber-stamped refactor can carry a behaviour change nobody asked for. So nothing it writes reaches this site until the release gate passes: lint, type check, tests, the AI-tell check and a full static build, all run by the deploy script before a single file uploads.
The money side: I pay for Max 5x, USD 100 a month since 10 Jul 2026, and I ran Max 20x before that. The smaller plan roughly halved my weekly hours on the top models, so repetitive bulk work now goes to cheaper tiers and Claude Code keeps the work that needs judgment.
The thing I want and do not have yet is first-class multi-repo awareness. Claude Code is excellent inside one repo; refactors that cross repos still need me to chain sessions by hand.
Run it. Gate it. Trust but verify.
Why it matters in 2026
Terminal agents now carry real multi-step work, not just autocomplete. Claude Code and OpenAI's Codex CLI are the two terminal agents I use, and subagents are what let a long, many-step job fit inside one working session.
What it costs
As of
Included with Claude Pro (USD 17/mo on annual billing, USD 20 month to month) and Claude Max (USD 100/mo for 5x, USD 200/mo for 20x). Not included on the Free plan. Or pay per token through the API. Checked 28 Sep 2026.
Use when
- +Daily code work where you want an agent in your terminal
- +Multi-step refactors with explicit plan-then-execute
- +Work across several repos where subagents keep the main session lean
- +Hard bugs requiring read-edit-run-iterate loops
Skip when
- xWhen you prefer IDE-pane integrations (Cursor, Cline win there)
- xPair-programming flows where you want suggestions, not actions
- xPure code-completion needs (Copilot is faster for that)
Hands-on guides
Alternatives I would consider
Read next
Adjacent in the playbook
As of 2026-09-28: Free tier with limits. Pro: Rs 1,920/mo (USD 20). Teams: Rs 3,840/seat/mo (USD 40).
Cursor, the IDE I Tried and the Empire's Soft Pass
As of 2026-09-28: Free and open source (Apache-2.0). API costs at vendor rate. Ollama-backed runs at zero marginal cost.
Cline, the VSCode Agent That Stays Out of My Way
As of 2026-09-28: Free and open source (Apache-2.0). API costs at vendor rate, and scale with how much you use it.
Aider, the Terminal Pair-Programmer With a Soul
As of 2026-09-28: Claude Pro: USD 17/mo on annual billing (USD 200 up front) or USD 20 month to month. Claude Max: USD 100/mo for 5x Pro usage, USD 200/mo for 20x (web prices). API per 1M tokens: Sonnet 5 USD 2 in / 10 out, Opus 5.5 USD 4 / 20, Fable 5.1 USD 10 / 50, Haiku 4.5 USD 1 / 5. Checked 28 Sep 2026.