Remote input and results
Send natural language or commands from WeChat, WeCom, or Telegram, run them in the local CLI, and receive results in the same conversation.
↔
WECHAT / WECOM / TELEGRAM × NATIVE LOCAL CLI
WeChat and local CLIs,
truly two-way.
More than forwarding a final answer. Inputs, approvals, questions, status, and sessions stay continuous between WeChat, WeCom, or Telegram and your local CLIs, while Codex, Claude Code, OpenCode, and Pi keep running natively on your computer.
✓ Local execution✓ Native terminals✓ Three channels
Sending taskRouting to the local CLILocal CLI is executingResults, approvals, and status are returning
Remote channels
Local workspace
—GitHub Stars
1.2.3Current version
—Total npm downloads
Quick start
Connect a local CLI in three steps.
Install, configure a channel, then choose either one direct launcher or the daemon. The launch commands are alternatives, not sequential steps.
01Install globally
$ npm install -g cli-wechat-bridge@latest02Log in / pair
$ wechat-setup03Choose a launcher
$ wechat-codexChannel
Launch mode
Requires Node.js ≥ 22.13.0 and at least one supported CLI. Pi 0.85.0+ requires Node.js ≥ 22.19.0. Telegram provides global commands since 1.2.3; send one WeChat message after launch to refresh the context token; Codex ≥ 0.162 can attach the current conversation.
Beyond final-answer forwarding
Two-way through the entire task.
Your phone is the remote entry point; the native terminal remains the primary workspace. Start from either side and continue in the same local session.
Phone and terminal continuity
The native CLI stays visible and usable, ready for you to continue from the current context when you return.
↔Approvals and questions
Route supported approval, clarification, and selection requests to your phone instead of waiting at the terminal.
↔Status and session control
Switch CLIs with /codex, /claude, /opencode, and /pi; inspect status, stop tasks, or resume sessions with /status, /stop, and /resume.
↔Two-layer architecture
Channels provide reach. The Bridge preserves native behavior.
Messages enter the current workspace through a daemon or direct launcher, then reach the matching adapter. Each adapter uses the CLI's native communication path instead of replacing it with another hosted chatbot. Regular Codex bridging supports 0.149.x–0.161.x; attaching the current conversation needs Codex 0.162+ with its shared daemon.
01 · Channel
WeChat iLink
WeCom Smart Bot
Telegram Bot API
02 · Workspace bridge
Daemon / direct launcher
Turn & session routing
Approvals / local companion
03 · Native CLI
Codex · WebSocket RPC
Claude Code · PTY + Hook
OpenCode · HTTP + SSE (1.18 / 2.0)
Pi · TUI extension
Trust and boundaries
A remote entry point, not a hosted workspace.
The Bridge runs on your computer with your local user permissions. It connects existing local CLIs without moving your project into a web workspace.
Review permissions and adapters ↗Work stays local
The workspace, CLI processes, and session state remain on your computer; channels carry messages and media.
Bound channel identity
Personal WeChat uses the logged-in account; WeCom pairs an operator with configured bot credentials; Telegram accepts only the paired operator.
Native permissions
Codex and Claude Code retain their CLI permission semantics. Pi runs as a fully approved local agent.
Visible boundaries
A daemon is bound to its startup directory, and you can observe, take over, or stop sessions locally.
Real runs
Diagrams explain the system. Real captures prove it works.
Existing captures remain as supporting evidence. Hover to pause and click to view the original.
