{"slug": "burnkit-stop-being-the-human-event-loop-for-your-ai-coding-sessions", "title": "BurnKit – Stop being the human event loop for your AI coding sessions", "summary": "BurnKit, a three-tool developer kit for running Claude Code and Codex in parallel, routes work to the appropriate provider, colors idle terminal tabs when AI is waiting for human input, and tracks plan burn before expensive windows expire. The tool exposes the workflow bottleneck around the human operator, turning hidden costs like provider choice, idle sessions, and context switching into signals that push developers toward building a real agent harness. BurnKit is designed as a pressure rig to squeeze idle time and manual babysitting, forcing the question of how to make AI sessions queue, split work, verify results, and ship without constant human intervention.", "body_md": "Overclock the human. Then build the harness.\n\nBurnKit is a three-tool kit for developers running Claude Code and Codex in parallel. It routes work to the right provider, colors idle terminal tabs when AI is waiting for you, and tracks plan burn before expensive windows quietly evaporate.\n\nChinese spirit name: `卷王三件套`\n\n.\n\nThis is not about going productivity-crazy, and it is not about making you go crazy either.\n\nBurnKit exposes the awkward truth inside AI coding: models keep getting faster, but the workflow still jams around the human operator. You think you need a stronger model. Then you notice the real drag: provider choice, idle sessions, context switching, wasted plan windows, and the person who keeps answering \"should I continue?\"\n\nBurnKit turns those hidden costs into signals. When the signals become too many to handle manually, the next question becomes obvious: how do you make AI ask less, queue work, split tasks, verify results, and ship without constant babysitting?\n\nThat is the harness entrance.\n\nTip:If your menu bar gets crowded, consider a free tool like[Dozer]to hide less-used icons and keep the Burn AI menubar signal front and center.\n\n```\nnpm install -g burnkit\nburnkit install all\n```\n\nOr run without installing:\n\n```\nnpx burnkit install all\n```\n\nFrom source:\n\n```\ngit clone https://github.com/hanzhangzzz/burnkit.git\ncd burnkit\nbin/burnkit install all\n```\n\nThen edit your provider config:\n\n```\n$EDITOR $(dirname $(readlink -f ~/.local/bin/c))/config.env\n```\n\nThe provider template lives at `tools/claude-provider-router/config.env.example`\n\n.\n\nRun Claude Code through the short router command:\n\n```\nc 0\nc team 2 0\n```\n\nCheck plan burn:\n\n```\nburnkit status --refresh\n```\n\nUninstall everything:\n\n```\nburnkit uninstall all\n```\n\nCopy this into your AI coding agent and let it handle the install:\n\n```\nInstall BurnKit for me in this repo.\n\nRules:\n- First run `scripts/e2e-install-verify.sh --dry-run`.\n- Run `bin/burnkit doctor`.\n- Dry-run each BurnKit install target before any real install:\n  - `bin/burnkit install router --dry-run`\n  - `bin/burnkit install tabs --dry-run --skip-python-check`\n  - `bin/burnkit install burn --dry-run`\n- Do not overwrite `tools/claude-provider-router/config.env`. If it already exists, preserve it byte-for-byte.\n- Before real install, tell me exactly what files and system state will be changed, then wait for my explicit confirmation.\n- After I confirm, run the real install and verify it with `scripts/e2e-install-verify.sh --real`.\n```\n\nThe important part: `config.env`\n\ncontains provider tokens. A correct agent must preserve it if it already exists.\n\n```\n1. Start more AI sessions.\n2. Watch idle tabs turn green, yellow, and red.\n3. Use Burn AI to see whether your 5h / 7d plan windows are being wasted.\n4. Hit the human scheduling limit.\n5. Start designing a real agent harness.\n```\n\nThis is the point. BurnKit is not a forever-babysitting tool. It is a pressure rig. It squeezes your idle time, your context switching, and your confidence that one human can keep ten agent sessions moving by hand.\n\nAt the limit, the useful questions stop being motivational and start being architectural:\n\n```\nWhy does it keep asking me?\nWhy can't it decide the next step?\nWhy am I still the human event loop?\nWhy can't these sessions queue, split work, verify, and ship?\n```\n\nExactly. You start harnessing.\n\n| Command | Purpose |\n|---|---|\n`burnkit doctor` |\nCheck local prerequisites and tool readiness |\n`burnkit install router` |\nRun the router internal installer; create `config.env` from the template only when missing, and install the `c` shim |\n`burnkit install tabs` |\nInstall iTerm2 Tab Color through the unified BurnKit entrypoint |\n`burnkit install burn` |\nInstall/build Burn AI, then run `burn-ai install` |\n`burnkit install all` |\nInstall CLI shim, router, tab color, and Burn AI in order |\n`burnkit uninstall all` |\nUninstall tab color, Burn AI, and CLI shim |\n`c 0` |\nStart Claude Code through provider config `0` |\n`c team 2 0` |\nStart Agent Team routing: leader on `2` , teammate on `0` |\n`burnkit router 0` |\nLong-form compatibility wrapper for `c 0` |\n`burnkit burn doctor` |\nForward to Burn AI CLI |\n`burnkit status --refresh` |\nRefresh and print plan usage state |\n\nEach tool still owns its own runtime files, safety checks, and uninstall path.\n\n| Color | Meaning | Operator Signal |\n|---|---|---|\n| Green | AI just finished and is waiting | Collect the result now |\n| Yellow | It has waited for a while | Your parallelism is leaking |\n| Red | It has waited too long | The machine is ready; the human is late |\n| White | Active tab, processing, or clean state | No attention needed here |\n\nOnly inactive tabs get colored. The tab you are looking at stays white because notifications should point at what you are missing.\n\n```\n.\n├── bin/\n│   └── burnkit\n├── tools/\n│   ├── claude-provider-router/\n│   ├── iterm2-tab-color/\n│   └── burn-ai/\n├── assets/\n├── package.json\n├── AGENTS.md\n├── CLAUDE.md\n├── README.md\n└── README.zh-CN.md\n```\n\n`tools/claude-provider-router/config.env`\n\ncontains tokens and must not be committed.- Burn AI does not manage login state, credentials, or private usage APIs. It reads local usage data already produced by Claude Code and Codex.\n- Burn AI does not overwrite existing Claude Code status line scripts. If one exists, it asks before installing a wrapper; skipping prints manual integration steps and explains which Claude features stay unavailable.\n- Tab color behavior, state cleanup, process detection, hook events, and daemon scheduling are behavior changes. Do not bundle them with docs or release polish.\n\nFor the release entry point:\n\n```\nbash -n bin/burnkit\nbin/burnkit --help\nbin/burnkit doctor\nscripts/e2e-install-verify.sh --dry-run\n```\n\nFor real install verification on a local machine:\n\n```\nscripts/e2e-install-verify.sh --real\n```\n\nThe e2e verifier checks both router install paths: missing `config.env`\n\nis created from the template with mode `600`\n\n, and an existing `tools/claude-provider-router/config.env`\n\nis preserved byte-for-byte with its original permissions.\n\nFor iTerm2 Tab Color changes:\n\n```\nbash -n tools/iterm2-tab-color/install-core.sh tools/iterm2-tab-color/uninstall-core.sh tools/iterm2-tab-color/install.sh tools/iterm2-tab-color/uninstall.sh tools/iterm2-tab-color/tab_color_hook.sh\npython3 -m py_compile tools/iterm2-tab-color/tab_color_daemon.py tools/iterm2-tab-color/reset_tab.py tools/iterm2-tab-color/test_daemon.py\npython3 -m unittest tools/iterm2-tab-color/test_daemon.py\n```\n\nFor Burn AI changes:\n\n```\ncd tools/burn-ai\nnpm ci\nnpm test\nnpm run build\nnpx --no-install burn-ai install\nburn-ai install\nnpx --no-install burn-ai doctor --dry-run\nnpx --no-install burn-ai status --fixtures\nnpx --no-install burn-ai menubar render\ngit diff --check\n```\n\nWhen installing Burn AI from this repository, prefer `bin/burnkit install burn`\n\nor `bin/burnkit install all`\n\n; those paths rebuild `tools/burn-ai`\n\nbefore copying the runtime into `~/.burn-ai/app`\n\n.\n\n|\n|", "url": "https://wpnews.pro/news/burnkit-stop-being-the-human-event-loop-for-your-ai-coding-sessions", "canonical_source": "https://github.com/hanzhangzzz/burnkit", "published_at": "2026-05-26 11:56:00+00:00", "updated_at": "2026-05-26 12:09:50.587661+00:00", "lang": "en", "topics": ["ai-tools", "ai-agents", "ai-products", "artificial-intelligence", "mlops"], "entities": ["BurnKit", "Claude Code", "Codex", "Dozer", "hanzhangzzz"], "alternates": {"html": "https://wpnews.pro/news/burnkit-stop-being-the-human-event-loop-for-your-ai-coding-sessions", "markdown": "https://wpnews.pro/news/burnkit-stop-being-the-human-event-loop-for-your-ai-coding-sessions.md", "text": "https://wpnews.pro/news/burnkit-stop-being-the-human-event-loop-for-your-ai-coding-sessions.txt", "jsonld": "https://wpnews.pro/news/burnkit-stop-being-the-human-event-loop-for-your-ai-coding-sessions.jsonld"}}