pikiclaw

Put the world's smartest AI agents in your pocket. Command local Claude, Codex & Gemini via best IM.

Let the best IM app become a top-tier Agent console on your computer

npx pikiclaw@latest

Demo

Real task: ask pikiclaw to gather and summarize today's AI news — the agent reads, writes, and sends results back through Telegram, all from your phone.

Basic operations: send a message, watch the agent stream, receive files back.

Why pikiclaw?

Most "IM + Agent" solutions either reinvent the agent (worse than official CLIs), run in remote sandboxes (not your environment), or only support short conversations (unusable for real tasks).

pikiclaw takes a different approach:

• Official Agent CLIs — Claude Code, Codex, Gemini CLI as-is, not a home-grown wrapper
• Your own machine — local files, local tools, local environment
• Your existing IM — Telegram, Feishu, or WeChat, no new app to learn

  You (Telegram / Feishu / WeChat)
          |
          v
       pikiclaw
          |
          v
  Claude Code / Codex / Gemini CLI
          |
          v
     Your Computer

It's designed for the moment you walk away from your desk — the agent keeps working locally, and you stay in control from your phone.

Quick Start

Prerequisites

• Node.js 18+
• At least one Agent CLI installed and logged in:
  • claude (Claude Code)
  • codex (Codex CLI)
  • gemini (Gemini CLI)
• A bot token for your IM channel (Telegram Bot Token, Feishu app credentials, or WeChat account)

Install & Launch

cd your-workspace
npx pikiclaw@latest

This opens the Web Dashboard at http://localhost:3939, where you can:

• Connect IM channels (Telegram / Feishu / WeChat)
• Configure agents and models
• Manage macOS system permissions
• Set up browser & desktop automation extensions
• Monitor sessions and system resources

npx pikiclaw@latest --setup   # interactive terminal wizard
npx pikiclaw@latest --doctor  # check environment only

Dashboard

IM Access — Telegram, Feishu, WeChat channel status and configuration

Agent Config — Default agent / model / reasoning effort, available agents overview

System Permissions — macOS accessibility, screen recording, disk access

Extensions — Managed browser & desktop automation (Appium Mac2)

Sessions — Per-agent session list and runtime status

System Info — Working directory, CPU / memory / disk monitoring

Features

Channels & Agents

• Telegram, Feishu, and WeChat — run one or all simultaneously
• Claude Code, Codex CLI, and Gemini CLI via unified driver registry
• Model listing, session management, and usage tracking through a single interface

Runtime

• Streaming preview with continuous message updates
• Session switching, resume, and multi-turn conversations
• Task queue with Steer — interrupt the running task and let a queued message jump ahead
• Working directory browsing and switching
• File attachments automatically enter the session workspace
• Long-task sleep prevention, watchdog, and auto-restart
• Long text auto-splitting; images and files sent back to IM directly
• Light / dark theme and i18n (Chinese & English)

Skills

• Project-level skills at .pikiclaw/skills/*/SKILL.md
• Compatible with .claude/commands/*.md
• Legacy .claude/skills / .agents/skills support with migration path
• Trigger via /skills and /sk_<name> in chat

Codex Human Loop

When Codex requests additional user input mid-task, pikiclaw surfaces the question as an interactive prompt in your IM. Reply there and the task continues.

MCP & GUI Automation

Each agent stream launches a session-scoped MCP bridge that injects local tools:

• im_list_files — list session workspace files
• im_send_file — send files back to IM in real time

Optional GUI capabilities:

• Browser automation — managed Chrome profile via @playwright/mcp; log in once, reuse across tasks
• macOS desktop automation — Appium Mac2 with desktop_open_app, desktop_snapshot, desktop_click, desktop_type, desktop_screenshot

Commands

| Command | Description | |---|---| | /start | Show entry info, current agent, working directory | | /sessions | View, switch, or create sessions | | /agents | Switch agent | | /models | View and switch model / reasoning effort | | /switch | Browse and switch working directory | | /status | Runtime status, tokens, usage, session info | | /host | Host CPU / memory / disk / battery | | /skills | Browse project skills | | /restart | Restart and re-launch bot | | /sk_<name> | Run a project skill |

Plain text messages are forwarded directly to the current agent.

Configuration

• Persistent config lives in ~/.pikiclaw/setting.json
• The Dashboard is the primary configuration interface

Browser automation is managed by the dashboard and runtime together — a dedicated Chrome profile is created and reused automatically. Just log in to the sites you need once in that browser.

macOS desktop automation requires Appium Mac2:

npm install -g appium
appium driver install mac2
appium

Then grant macOS Accessibility permission to your terminal app.

Relevant environment variables:

• PIKICLAW_DESKTOP_GUI
• PIKICLAW_DESKTOP_APPIUM_URL

Roadmap

• ACP (Agent Client Protocol) adoption — unified driver for any ACP-compatible agent, replacing per-agent CLI output parsing. See ACP Migration Plan
• Expand session-scoped MCP bridge into a more complete top-level tool layer
• Improve GUI automation, especially browser + desktop tool coordination
• More IM channels (WhatsApp, etc.)

Development

git clone https://github.com/xiaotonng/pikiclaw.git
cd pikiclaw
npm install
npm run build
npm test

npm run dev          # local dev (--no-daemon, logs to ~/.pikiclaw/dev/dev.log)
npm run build        # production build
npm test             # unit tests
npm run test:e2e     # end-to-end tests
npx pikiclaw@latest --doctor  # environment check

See also: ARCHITECTURE.md · INTEGRATION.md · TESTING.md

License

MIT