AutoKaam Playbook
Claude Code, the CLI Agent I Run All Day
Anthropic's terminal-first coding agent; the empire's daily driver for code work.
Last reviewed:
The operator take
Claude Code is the empire's primary coding interface. I run version 2.1.131 across every empire repo (autokaam, papersadda, taxwallaai, kaam-tracker, recon, ghatak-prachar) and it is the difference between solo founder and feeling-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 spawn its own subagents for parallel work (Leg A through Leg G in this very session), run shell commands with appropriate permission gates, and ship clean diffs.
The 2.1 series shipped subagent forking, which is the killer feature for empire work. I can fan out 6 parallel work streams from one orchestrator session, and the parent only consumes the summary, not the full transcript. That changed how I scope big migrations.
What I have screwed up: trusting it on first-pass refactors without reviewing the diff. Claude Code is excellent but not infallible, and on three occasions it has applied a refactor that introduced subtle behaviour changes I missed because I rubber-stamped. Now I gate all its commits through a pre-commit hook that runs my AI-tell gate, my lint suite, and my type checker. The hook catches drift before it ships.
The Indian-operator angle is the Max OAuth flat rate. At Rs 8,500 a month for Max, Claude Code's per-session quota stretches across roughly 30-40 productive hours of agentic work for me. That works out to under Rs 300 an hour for an agent that handles tasks an Indian senior engineer would charge Rs 2,500 an hour for. The leverage is asymmetric.
The thing I want and do not have yet is first-class multi-repo awareness. Claude Code is excellent within a repo; cross-repo refactors (the empire has 20-something repos) still need me to chain it manually.
Run it. Pre-commit-hook it. Trust but verify. This is the empire pattern.
Why it matters in 2026
Agentic coding is the dominant developer interface in 2026. Claude Code is one of two terminal-first agents (alongside Codex CLI) that have crossed the production-quality threshold. The subagent forking primitive is unique and load-bearing for serious multi-step work.
Cost in INR
Free tier available with Claude Pro. Claude Max (recommended for empire-scale work): ~Rs 8,500/mo. Pay-as-you-go API also supported.
Use when
- +Daily code work where you want an agent in your terminal
- +Multi-step refactors with explicit plan-then-execute
- +Empire-style multi-repo work where subagent forking helps
- +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)
Alternatives I would consider
Read next
Adjacent in the playbook
Free tier with limits. Pro: Rs 1,700/mo (USD 20). Business: Rs 3,400/seat/mo (USD 40).
Cursor, the IDE I Tried and the Empire's Soft Pass
Free extension. API costs at vendor rate. Ollama-backed runs at zero marginal cost.
Cline, the VSCode Agent That Stays Out of My Way
Free open source. API costs at vendor rate. Sessions average Rs 200-400 with Sonnet.
Aider, the Terminal Pair-Programmer With a Soul
Claude.ai Pro: ~Rs 1,700/mo (USD 20). Claude Max (5h sessions): ~Rs 8,500/mo (USD 100). API Sonnet 4.6: USD 3 input / USD 15 output per 1M tokens. API Opus 4.7: USD 15 input / USD 75 output per 1M tokens.