{"slug": "show-hn-a-terminal-where-ai-coding-clis-can-mention-each-other", "title": "Show HN: A terminal where AI coding CLIs can mention each other", "summary": "SHIKISHA-TERM, a terminal application from developer styleio, launched on the Microsoft Store and GitHub Releases, letting multiple AI coding CLIs run in one window and hand work to each other via @mentions. The tool supports Claude Code, Codex CLI, Gemini CLI, Aider, DeepSeek, Ollama, LM Studio and OpenRouter, runs agents locally or on cloud MicroVMs, and can be steered from a phone via QR code; it stores no API keys and uses existing AI subscriptions. Linux builds ship as a headless component reachable from a browser or phone, with Debian/Ubuntu and Fedora/RHEL packages plus a static binary.", "body_md": "**Run many AI agents at once and let them @mention each other for help — on your PC, in the cloud, or from your phone.**\n\n  For Windows, from the Microsoft Store or as a portable zip. Uses the AI subscriptions you already have.\n\n**Works with** Claude Code · Codex · Gemini CLI · Aider · OpenAI · Anthropic · DeepSeek · Mistral · OpenRouter · Ollama · LM Studio — or any shell over SSH\n\n<sub>⭐ If this is useful, a **star** helps other people find it.</sub>\n\n**@mention one agent from another, like in a chat app.**\n“Build this feature, ask @codex to review it, and loop until there is nothing left to fix.”\nThat one line starts the write-and-review loop. No more being the copy-paste courier.\n\n**Check on your AI from your phone — and tell it to carry on.**\nScan a QR code and you can see what every tab is doing and send instructions,\nfrom the train, from a café, from bed.\n\nRunning one terminal AI is easy. Running **four** is not.\n\nYou end up with a window per agent, no idea which one is waiting for you, and a lot of\ncopy-pasting between them. SHIKISHA-TERM puts them in one window and knows the difference\nbetween *working*, *finished* and *waiting for a human*. Agents hand work to each other with an\n`@` mention and read the answer back. They can run on your PC or on a cloud MicroVM, and you can\nsteer them from your phone.\n\nIt talks to whatever runs in a terminal, so it is not tied to one vendor: Claude Code, Codex CLI, Gemini CLI, DeepSeek, Ollama, Aider, or a plain shell over SSH.\n\n**Most AI terminals run a single agent. This one runs several — on your PC or in the cloud.**\n\nFor example:\n\n- **“Ask @codex to review it, and loop until it is clean”** — one line, and two agents go back and forth.\n- **Four agents on four branches** , each in a worktree of its own, all in one window.\n- An agent in **YOLO mode on a cloud MicroVM** , sealed off from your PC.\n- An AI that opens your dev server in the **built-in browser** and debugs it on its own.\n- Kick off a long job and **check on it from your phone** on the way home — and tell it to carry on.\n\n**From the Microsoft Store** — [get it here](https://apps.microsoft.com/detail/9PB8XQVM87Z0). Microsoft signs the Store copy, so\nthere is no security warning to click past, and updates arrive on their own.\n\n**Or take the portable copy**, if you would rather nothing were installed:\n\n1. Download `SHIKISHA-TERM.zip` from[Releases](https://github.com/styleio/ShikishaTerm/releases/latest)\n2. Unzip it anywhere — a USB stick and a synced folder both work\n3. Run `SHIKISHA-TERM.exe`\n\nIt opens a window of its own. Nothing needs to be running it — no terminal to set up, no font or colour scheme to get right first.\n\nNothing is written outside that folder. There is no installer and no runtime to install.\n\nIt uses **your existing AI subscriptions** through their CLIs. No API keys are stored,\nand none are needed.\n\n**There is no window on Linux.** What runs there is the part that does the work — it opens\nthe terminals, watches what each agent is doing, runs your automation — and you reach it\nfrom a browser or a phone. Put it on a VPS or a cloud VM and the agents keep going when\nyour laptop is shut.\n\n**Debian / Ubuntu**\n\n```\ncurl -fsSL https://pkg.shikisha-term.com/shikisha-archive-key.asc \\\n  | sudo tee /etc/apt/keyrings/shikisha.asc >/dev/null\necho \"deb [signed-by=/etc/apt/keyrings/shikisha.asc] https://pkg.shikisha-term.com/apt stable main\" \\\n  | sudo tee /etc/apt/sources.list.d/shikisha.list\nsudo apt update && sudo apt install shikisha\n```\n\n**Fedora / RHEL / openSUSE**\n\n``` python\nsudo rpm --import https://pkg.shikisha-term.com/shikisha-archive-key.asc\nsudo tee /etc/yum.repos.d/shikisha.repo <<'REPO'\n[shikisha]\nname=SHIKISHA-TERM\nbaseurl=https://pkg.shikisha-term.com/rpm\nenabled=1\ngpgcheck=1\nrepo_gpgcheck=1\ngpgkey=https://pkg.shikisha-term.com/shikisha-archive-key.asc\nREPO\nsudo dnf install shikisha\n```\n\n**Anything else** — one static binary, into your own home folder, asking for nothing:\n\n```\ncurl -fsSL https://raw.githubusercontent.com/styleio/ShikishaTerm/main/packaging/linux/install.sh | sh\n```\n\nInstalling does not start anything. When you want it running:\n\n```\nsystemctl --user enable --now shikisha\nloginctl enable-linger \"$USER\"   # keep it running after you log out\n```\n\nIt prints the address to open the board at. Everything is signed with\n[one key](https://pkg.shikisha-term.com/shikisha-archive-key.asc), and every route above\nchecks it before installing.\n\n**The Store copy is signed by Microsoft and shows no warning.** This section is about the\nzip only.\n\nThe zip carries no code-signing certificate, so the first run of an unzipped copy shows\n**\"Windows protected your PC\"** ([code signing policy](https://github.com/styleio/ShikishaTerm/blob/main/SIGNING.md)). Click\n**More info → Run anyway** to continue.\n\n**You can help make that warning go away.** Removing it means a certificate, and\n[SignPath Foundation](https://signpath.org/) issues them to open source projects for free —\nbut for a program like this one they ask for *\"a certain verifiable reputation\"* before\nputting their own name behind it. For a project on GitHub the most legible form that\nreputation takes is stars. So if this is useful to you,\n**[star the repository](https://github.com/styleio/ShikishaTerm)** — it takes a second, and it is the thing most likely to\nget a signed build into the hands of whoever downloads the zip after you.\n\nIf you would rather check the download first — and being wary of an unsigned executable from the internet is the right instinct:\n\n- \nEvery release is built by [GitHub Actions](https://github.com/styleio/ShikishaTerm/blob/main/.github/workflows/release.yml) from this\nsource, not on anyone's laptop. The build log for your download is public\n- \nA `SHIKISHA-TERM.zip.sha256` is published next to the zip:\n\n```\nGet-FileHash SHIKISHA-TERM.zip -Algorithm SHA256\n```\n\nThe first run opens **Getting set up**. Pick your main AI from the ones installed on this PC and\ncheck GitHub CLI; an AI you do not have yet opens its install page right there.\n\nThen the screen points you to **add a project**: pick a folder or clone from a Git URL (onto a\nMicroVM or an SSH server too). Press where it points next, and the AI starts in that folder —\nno JSON editing required.\n\nTo open just the settings (also the way back in if a broken config stops the app from\nstarting), double-click **`Settings.cmd`**.\n\n- **Call one with `@`, like in chat** — type`@` in the input bar to list this desk's tabs; a pick\nbecomes a chip. The agent in front hands the work over, reads the answer and carries on\n- **Review loops** — “ask @codex to review it, and loop until there is nothing left to fix” runs\nthe back-and-forth. The app counts the rounds, up to the limit under Settings > “Limits on\nhanding work”\n- **Terminals and pages too** — run a command in a named terminal, or give a named web page a goal\nin plain words. Only the tabs a person named with`@` can be driven\n- **Nobody waits on a long job** — its answer comes back to the tab that asked when it is done\n\n- **Tabs with real status** — every tab shows whether it is working, done, or waiting for\nyou, detected from the screen itself rather than from any vendor API\n- **Split panes** — split as many ways as you like: agents side by side, or an agent next to\nthe browser it is driving\n- **Worktrees** — a folder of its own for every branch, so several branches move at once\n- **Pick up where you left off** — close the app and resume the conversation (Claude Code, Codex)\n- **Notifications** — Slack / Discord / Telegram, this PC, or your phone when a job finishes\n\n- **A machine of its own** — clone the project onto a cloud MicroVM (E2B) and run the agents\nthere. Whatever they do never reaches your PC, which makes it the place for YOLO mode\n- **Set up once** — every worktree is a copy of the project's machine, so what you installed\nand signed in to is already there\n- **Pauses when idle** — it keeps everything and picks up where it stopped when its URL or a\nwebhook is called\n- **The token stays outside** — the git server's token is added on the way out, so no agent\ninside can read it\n- **Previews by URL** — open what the machine serves by URL, unlisted or private\n\n- **Browser** — open pages inside the app and let an AI drive and check them. A dev server's\nport opens in a browser tab too\n- **Git panel** — stage part of a diff, let an AI write the commit message, and let an AI tab\nwork through a conflict\n- **SFTP** — this PC on the left, the server on the right; send and fetch files\n- **A real terminal** — SSH, Docker, WSL, jump hosts, key files, port forwarding, session\nlogs, legacy encodings, IME input and mouse support\n\n- Scan a QR code to see every tab and send instructions. Everything you can do at the PC, you can do from the phone (see below)\n\n- **Desks** — swap the whole tab layout per project, like virtual desktops. Export one\nto a single file, automation scripts included, and hand it to another machine or person\n- **Automation** — \"when this finishes, hand the result to the review tab\", \"answer this\nconfirmation automatically\". Written in Lua, or described in plain language and written\nfor you by an AI you already have installed\n- **Runaway protection** — a limit on how many times agents may hand work to each other,\nan emergency stop, and per-tab input locks\n\nTurn on \"Use from your phone\" in the settings and a QR code appears. Scan it and you get a\nsmall web page listing every tab, its status, and a box to send instructions. `Ctrl+B 0`\nthen `i` shows the same code on the board.\n\nUnder the code sit the only two things worth acting on. **📋 copies the link** — token and\nall — to the clipboard, for a phone that cannot photograph the screen it is on, and the\n**badge beside it says which network that link leads to**: green for Tailscale (your own\ndevices, anywhere), amber for the home LAN (everyone on that Wi-Fi). The link itself is\nnever printed: with its token it is the key to the machine, and without one it opens\nnothing.\n\n**Know where it is reachable from** — this feature lets a phone run commands on your machine:\n\n- Only people **on the same network** can connect. It is never published to the internet\n- With **[Tailscale](https://tailscale.com/)** (free) only your own devices can reach it,\nencrypted, from anywhere. This is the safest way and the one to prefer\n- Without Tailscale it works on your **home LAN only** . Anyone on the same Wi-Fi who knows\nthe URL and the token could use it — do not turn it on for shared or public Wi-Fi\n- Binding to a public address (`remote.allow_public` ) never happens unless you write it in\nthe config file yourself\n\nSee [SECURITY.md](https://github.com/styleio/ShikishaTerm/blob/main/SECURITY.md) for the full threat model.\n\nThe prefix is `Ctrl+B`, tmux-style. `Ctrl+B ?` opens the help.\n\n| Key | Action | \n|---|---|\n| `Ctrl+B q` | Quit | \n| `Ctrl+B 0` –`9` | Switch tab (0 = INDEX) | \n| `Ctrl+B w` /`W` | Desk list / next | \n| `Ctrl+B %` /`\"` | Split the view beside / below ( `<``>` resize) | \n| `Ctrl+B o` / arrows | Move between panes | \n| `Ctrl+B X` | Close this pane (the tab keeps running) | \n| `Ctrl+B =` | Put every divider back to even halves | \n| `Ctrl+B l` | Toggle input lock | \n| `Ctrl+B r` /`R` | Restart the tab, carrying its conversation over / starting a new one. **On a tab you have not spoken to yet, it picks up the conversation from the last run of the app** | \n| `Ctrl+B [` | Copy mode ( `c` copies the latest response).`/` searches the history,`n` /`N` step through the matches | \n| `Ctrl+B a` /`x` | Automation on/off / emergency stop | \n| `Ctrl+B 0` →`i` | Show the QR code for phone access | \n| `Ctrl+B 0` →`k` | Set or change the master password (encrypts `secrets.json` ) | \n\nAn `on_notify` automation hook catches every terminal notification (bell/OSC, even over ssh),\nso you can forward it to your phone or route it — see docs/AUTOMATION.md.\n\n`browser_snapshot` captures a browser page as a PNG (see it under Settings → Snapshots) — a\nrally can keep a visual record of what it did.\n\n`Ctrl+B :` opens a command palette — one box to find and run anything: a tab, a setting,\nthe Vault, any key action, your own quick actions.\n\nThe board's \"Find\" searches both the live output of every open tab and every conversation the AI CLIs have recorded, by what was said in them, and reopening one adds a tab that resumes it in its folder.\n\nThe board's SESSIONS table shows a COST column — processor and memory per agent, summed over everything it started — so you can see which of several running agents is heavy without leaving the terminal.\n\nA browser login can be saved once (`browser_state_save`) and reloaded by a later rally\n(`browser_state_load`), so a rally does not start by logging in. It saves our own browser\nprofile's cookies, httpOnly and all; it does not crack Chrome's or Firefox's stores.\n\nEach tab says where it is, under its name: the branch, its pull request, and any ports it\nis listening on. Pull request numbers use the GitHub sign-in already on this PC (`gh auth login`, or `GITHUB_TOKEN`); without one the rest still shows.\n\nEvery key can be moved, including the prefix — the settings list each action with the\nkey it answers to now, and the help screen (`Ctrl+B ?`) shows what you actually have\nrather than what shipped.\n\nColours are chosen in the settings, by name. The list is every scheme this PC already\nhas — whatever Windows Terminal is carrying, plus any scheme file you drop into\n`config/themes`, in the format colour schemes are published in. Pick a light one and the\nwhole window follows, not only the terminal.\n\nThe mouse works too: wheel to scroll, **Ctrl+wheel to change the terminal's size**, drag to\ncopy, right click to paste, click a tab name to switch. Once the window is divided, each pane's caption carries ▥ and ▤ to divide it\nagain and ✕ to close that view, and the divider between panes can be dragged to rebalance\nthem — double-click it for even halves. The tab bar on the left works the same way: drag its\nright edge to set its width, double-click for the width it ships with, and drag it shut to\nput it away and give the whole window to the terminal. `Ctrl+B s` does that too, and brings\nit back the width it was.\n\nAn automation is a few lines of Lua dropped into a folder, named after the moment it should run:\n\n```\n-- on_done.lua — hand the finished work to the review tab, and stop after 5 rounds\nif tab.chain_depth == 0 then return end          -- a human started this; do nothing\n\nlocal rounds = shikisha.get_var(\"rounds\") or 0\nif tab.output:match(\"LGTM\") or rounds >= 5 then\n  shikisha.notify(\"slack\", \"Review finished (\" .. rounds .. \" rounds)\")\n  return\nend\nshikisha.set_var(\"rounds\", rounds + 1)\nshikisha.send_to_tab(\"reviewer\", \"Please fix these points:\\n\" .. tab.output)\n```\n\nAn agent can also say what it is doing — `shikisha.set_status(\"build\", \"running tests\")`\nputs that under its name in the tab bar, beside the state dot that is read off its screen.\nOne is what we could tell; the other is what it chose to say.\n\nAutomations run in a sandbox: **no file access and no network by default**, and\nnotifications can only go to targets you registered.\n\nThe same commands can be called **from outside the app**, over a named pipe, by anything\nthis app started — so a CLI sitting in a tab can split a pane, drive the browser, or hand\nwork to another tab without a human at the keyboard. Same names as in Lua, nothing new to\nlearn. See section 7 of the reference.\n\nFull reference: **[docs/AUTOMATION.md](https://github.com/styleio/ShikishaTerm/blob/main/docs/AUTOMATION.md)** — also reachable from the\nsettings screen.\n\n```\nSHIKISHA-TERM.exe   the app\nSettings.cmd           opens only the settings screen\nconfig.json            general settings + the list of desks\nsecrets.json           notification targets and tokens (encryptable, never share)\ndesks/            desk definitions (shareable per project)\nprofiles/              per-AI status detection rules\nscripts/               automation scripts\nlogs/                  session and automation logs\nlang/                  interface languages (en.json is the base)\ndocs/AUTOMATION.md     how to write automations\n```\n\nCopy `config.example.json` and the other `*.example.json` files to get started.\n\nThe interface follows your Windows language and falls back to English. You can also set it\nin the settings screen or with `\"language\"` in `config.json`.\n\n| Language | Status | \n|---|---|\n| English | ✅ base | \n| 日本語 | ✅ complete | \n| *your language* | [contributions welcome](https://github.com/styleio/ShikishaTerm/blob/main/docs/TRANSLATING.md) | \n\nCopy `lang/en.json` to `lang/<code>.json` and translate the values — that is the whole\nprocess. Keys you leave out fall back to English, so a partial translation still works.\nSee [docs/TRANSLATING.md](https://github.com/styleio/ShikishaTerm/blob/main/docs/TRANSLATING.md).\n\nIt is deliberately narrow, and honest about it:\n\n- **Windows only** today. It is built on ConPTY; other platforms are not supported yet —[★ star the repo](https://github.com/styleio/ShikishaTerm) to hear if that changes\n- **It drives CLIs, it does not replace them.** Your AI subscriptions, logins and settings\nstay exactly as they are\n- **It is a terminal, not an IDE.** It manages sessions and moves text between them\n\nTranslations, profiles for AI CLIs it does not know yet, bug reports — all welcome.\nStart with [CONTRIBUTING.md](https://github.com/styleio/ShikishaTerm/blob/main/CONTRIBUTING.md), and please keep to the\n[Code of Conduct](https://github.com/styleio/ShikishaTerm/blob/main/CODE_OF_CONDUCT.md). Questions or ideas? Open a\n[Discussion](https://github.com/styleio/ShikishaTerm/discussions).\n\n```\ncargo build --release\n```\n\nRust (MSVC toolchain) is required. The result is a single executable with no runtime dependencies.\n\n- [Automation reference](https://github.com/styleio/ShikishaTerm/blob/main/docs/AUTOMATION.md) — events, variables, commands, examples\n- [Translating](https://github.com/styleio/ShikishaTerm/blob/main/docs/TRANSLATING.md) — how to add a language\n- [Design notes](https://github.com/styleio/ShikishaTerm/blob/main/DESIGN.md) — terminology, architecture, and the safety model ([日本語](https://github.com/styleio/ShikishaTerm/blob/main/DESIGN.ja.md) )\n- [Security](https://github.com/styleio/ShikishaTerm/blob/main/SECURITY.md) — threat model and how to report an issue\n- [Code signing policy](https://github.com/styleio/ShikishaTerm/blob/main/SIGNING.md) — how release binaries are built and signed\n\nIf SHIKISHA-TERM saves you time, you can support its development on\n[Ko-fi](https://ko-fi.com/styleio). Thank you.\n\nThe program is built from open source components, and the terms of every one\nof them are reproduced in [THIRD-PARTY-NOTICES.txt](https://github.com/styleio/ShikishaTerm/blob/main/THIRD-PARTY-NOTICES.txt),\nwhich is inside the download as well as here.", "url": "https://wpnews.pro/news/show-hn-a-terminal-where-ai-coding-clis-can-mention-each-other", "canonical_source": "https://github.com/styleio/ShikishaTerm", "published_at": "2026-09-29 11:42:59+00:00", "updated_at": "2026-09-29 11:47:44.713018+00:00", "lang": "en", "topics": ["ai-agents", "ai-tools", "developer-tools", "ai-products"], "entities": ["SHIKISHA-TERM", "styleio", "Claude Code", "Codex CLI", "Gemini CLI", "Aider", "Ollama", "Microsoft Store"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/show-hn-a-terminal-where-ai-coding-clis-can-mention-each-other", "markdown": "https://wpnews.pro/news/show-hn-a-terminal-where-ai-coding-clis-can-mention-each-other.md", "text": "https://wpnews.pro/news/show-hn-a-terminal-where-ai-coding-clis-can-mention-each-other.txt", "jsonld": "https://wpnews.pro/news/show-hn-a-terminal-where-ai-coding-clis-can-mention-each-other.jsonld"}}