[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-81151":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":11,"openIssues":13,"contributorsCount":13,"subscribersCount":13,"size":13,"stars1d":13,"stars7d":13,"stars30d":13,"stars90d":13,"forks30d":13,"starsTrendScore":13,"compositeScore":14,"rankGlobal":9,"rankLanguage":9,"license":15,"archived":16,"fork":16,"defaultBranch":17,"hasWiki":18,"hasPages":16,"topics":19,"createdAt":9,"pushedAt":9,"updatedAt":20,"readmeContent":21,"aiSummary":22,"trendingCount":13,"starSnapshotCount":13,"syncStatus":23,"lastSyncTime":24,"discoverSource":25},81151,"teleport","thith\u002Fteleport","thith","Telegram bridge for AI coding agents (Claude Code, Codex, Gemini) — short reports out, short instructions back, no chat mirroring",null,"JavaScript",30,16,0,3.69,"MIT License",false,"main",true,[],"2026-06-12 02:04:11","# Teleport — Telegram bridge for AI coding agents\n\nA small tool that lets AI agents (Claude Code, Codex, Gemini, …) send short progress reports to your Telegram and take short instructions back, while you are away from the desk.\n\nShort reports, short replies — not a full session mirror.\n\nExtracted from internal tooling built for **[trumviahe.com](https:\u002F\u002Ftrumviahe.com)**.\n\n## Quick start\n\n### 1. Clone next to your projects\n\n```bash\ncd ~\u002FProjects   # any parent folder works (e.g. ~\u002Fwork, ~\u002Fcode, D:\\dev)\ngit clone https:\u002F\u002Fgithub.com\u002Fthith\u002Fteleport.git\ncd teleport\ncp .env.example .env\n```\n\n`teleport\u002F` must sit as a sibling of every project that uses it:\n\n```\n\u003Cyour-parent-folder>\u002F   # e.g. ~\u002FProjects, ~\u002Fwork, D:\\dev\n├── teleport\u002F\n├── ProjectA\u002F\n└── ProjectB\u002F\n```\n\nThe parent folder name doesn't matter — what matters is that `teleport\u002F` is a **sibling** of your projects.\n\n### 2. Get a bot token and your chat ID\n\n- **Bot token:** open [@BotFather](https:\u002F\u002Ft.me\u002FBotFather) on Telegram → `\u002Fnewbot` → follow prompts → copy the token.\n- **Personal chat ID:** open [@userinfobot](https:\u002F\u002Ft.me\u002Fuserinfobot) → press *Start* → copy your numeric ID.\n- **For a team group chat:** create a group, add your bot + people + [@RawDataBot](https:\u002F\u002Ft.me\u002FRawDataBot) → RawDataBot instantly prints the group ID (a negative number) → remove RawDataBot when done.\n- **(Optional) Multiple projects → enable Topics:** if you'll run teleport from more than one project, enable forum topics so each project lands in its own thread. In [@BotFather](https:\u002F\u002Ft.me\u002FBotFather): `\u002Fmybots` → select your bot → `Bot Settings` → `Topics` → *Enable*. Teleport auto-creates one topic per project (named after the cwd basename) on first send. No further setup; if Topics stays off, sends still work and just share one thread.\n\n### 3. Fill `.env`\n\n```ini\nREPORT_BOT_TOKEN=\u003Cbot token from BotFather>\nTELEGRAM_ADMIN_CHAT_ID=\u003Cyour chat ID or group ID>\n```\n\nThen verify it works:\n\n```bash\nnode scripts\u002Fsend-telegram.mjs \"hello from teleport\"\n```\n\nYou should see the message land in your Telegram. If not, fix `.env` before continuing.\n\n### 4. Wire it into your AI agents (one-time, global)\n\nOpen Claude Code, Codex, or Gemini CLI **inside any project sibling of `teleport\u002F`** (so `..\u002Fteleport\u002F` resolves correctly), then paste this prompt — the agent does the rest:\n\n> Enable Telegram reporting **globally for all my AI agents** by following `..\u002Fteleport\u002FREADME.md`. Copy the canonical snippet **verbatim** (do not paraphrase or rewrite) from the \"Wiring snippet\" section of that README into **all three** global config files so it works whichever agent I run later — `~\u002F.claude\u002FCLAUDE.md` (Claude Code), `~\u002F.gemini\u002FGEMINI.md` (Gemini), `~\u002F.codex\u002FAGENTS.md` (Codex). Create any file that does not exist yet. On Windows use the equivalent under `%USERPROFILE%`. After writing, send a test \"hello\" from whichever agent I am running now.\n\nThis wires Telegram support into **all three** global agent config files at once, so every project — and every agent you switch to later — picks it up automatically.\n\nYou can also do this manually if you wish: copy the snippet below into each of those files yourself. For per-project setup, write it into the project's local `CLAUDE.md` \u002F `GEMINI.md` \u002F `AGENTS.md` instead — useful if you only want Telegram on specific projects.\n\n##### Other agents\n\n- **Antigravity:** uses `~\u002F.gemini\u002FGEMINI.md` — already covered by the step above.\n- **Cursor:** no home directory file. Open Cursor → Settings → Rules → User Rules, and paste the canonical snippet (below) into the User Rules box.\n- **OpenCode, or any other agent:** start the agent, then paste this short prompt — it will read this README and configure itself: *\"Configure Telegram reporting for yourself by following `..\u002Fteleport\u002FREADME.md`. Find the 'Wiring snippet' section and copy it verbatim into your global config file (you know where it lives on this OS).\"*\n\n#### Wiring snippet (for the agent — or for manual install)\n\nThis is what the agent copies into each config file when you paste the prompt above. You normally do not need to read or touch this — it is shown here so the agent can find it, and so you can copy it yourself if you prefer to install manually.\n\n````markdown\n## Telegram Reporting\n\n**WHENEVER** the user asks to \"send a Telegram report\" (or variants: \"send via tele\", \"tele me\", \"ping me when done\", \"gửi tele\", \"báo cáo qua tele\", \"tele cho tôi khi xong\"…), you **MUST** read `..\u002Fteleport\u002Frules\u002Ftelegram-guide.md` and follow it. Look up your identity prefix in the guide's prefix table.\n\nScripts + guide are centralized at `..\u002Fteleport\u002F` (sibling of every project). Each project keeps no local copy. Do **NOT** rely on any invocation details from memory or from this config snippet; the guide is the single source of truth and may change as Teleport is updated.\n\nAfter sending any Telegram message, you **MUST immediately** start the reply-listener flow described in the guide's \"Listening for Replies\" section. **MUST NOT** skip or defer this step even if the task feels complete — the user may reply via Telegram at any time.\n````\n\n## How to use\n\nSay *\"tele me\"* (or similar) **before** you leave the desk. Examples:\n\n- *\"Ping me on Telegram when done.\"*\n- *\"Report via Telegram and wait for instructions.\"*\n- *\"Schedule a wakeup in 30 minutes; tele me when you wake up.\"*\n- *\"Send a tele report after each PR you open.\"*\n\nThe agent sends a short report, listens for your reply, and continues while you're away. *If the bot reacts 💔 on your message, it means it was ignored — you sent a plain message instead of using Telegram's Reply feature.*\n\n## Before you go away from the desk\n\nTeleport does not bridge permission dialogs. If the agent stops to ask \"may I run this?\", nobody on Telegram can answer and the agent stalls. Put the agent in a mode that runs without confirmation:\n\n- **Claude Code 4.7+:** Auto Mode (not `auto-accept` — they are different modes).\n- **Codex:** Auto-Review.\n- **Gemini CLI:** YOLO (`-y`).\n- **Antigravity:** Settings → **Auto Execution** (or *Terminal Command Auto Execution*) → choose **Always Proceed**. Also enable **Agent Non-Workspace File Access** — Teleport scripts live in `..\u002Fteleport\u002F`, outside the current workspace. If commands still get blocked, install the [YoloMode extension](https:\u002F\u002Fmarketplace.visualstudio.com\u002Fitems?itemName=mrkeles61.yolomode).\n- **Cursor:** Settings → Features → Agent → enable **YOLO mode** (auto-runs terminal commands without confirmation).\n- **Other agents (OpenCode, etc.):** any mode equivalent to \"run without confirmation\".\n\nFor older Claude Code \u002F Codex versions: use `--dangerously-skip-permissions` or Full Access mode.\n\nAlso: keep your host machine awake. The agent and the reply listener live on the same machine — if the OS sleeps, both die.\n\n- **macOS:** `caffeinate -i` in a terminal for the session.\n- **Linux desktop:** `systemd-inhibit --what=sleep -- sleep infinity`.\n- **Windows:** set the active power plan's sleep timer to *Never*.\n\nTeleport can also run on a VPS \u002F cloud VM (no host-awake needed there).\n\n## Why not \u002Fremote-control or other Telegram mirrors?\n\nThose tools mirror your full local session to your phone — every thinking token, tool call, and partial diff. On a small screen you end up scrolling and micro-managing every line, which defeats the point of being away from the desk.\n\nTeleport does the opposite: full conversation stays on your laptop; only short reports go to your phone, only short replies come back. Like a human assistant in the next room. If you don't trust the agent to run unattended, use `\u002Fremote-control` instead — Teleport is the wrong tool for that.\n\n## Notes\n\n- **Zero deps:** pure Node built-ins.\n- **Requirements:** `node` on `PATH`; every project sits as a sibling of `teleport\u002F`; one bot, one chat, one shared `.env`.\n- **Don't rely on it for critical work.** Agents sometimes forget to send follow-up messages.\n","Teleport 是一个用于AI编码助手（如Claude Code、Codex、Gemini等）与Telegram通信的小工具，支持发送简短进度报告并接收简短指令。其核心功能在于为用户提供了一种在离开办公桌时仍能监控和指导AI工作的便捷方式，而不进行完整的会话镜像。项目采用JavaScript编写，通过简单的环境配置即可将Telegram机器人与用户的AI代理集成起来，适用于需要远程管理多个开发项目的场景。无论是个人开发者还是团队协作，只要按照文档指引设置好Telegram Bot令牌和个人或群组聊天ID，并将相关代码片段添加到各AI代理的全局配置文件中，就可以轻松启用这一功能。",2,"2026-06-11 04:03:42","CREATED_QUERY"]