HelloClaw

An AI-powered developer assistant with CLI and Telegram interfaces for task automation, strategic planning, Q&A, web scraping, and approval-based agent execution.

Full write-up:

HelloClaw is an AI-powered developer assistant designed to combine task automation, planning, research, and interactive AI assistance into a unified CLI and Telegram-based interface. The project uses OpenRouter for LLM access and provides multiple operational modes for different types of workflows.

The CLI provides three core modes: Agent, Plan, and Ask. Agent Mode focuses on automated task execution, where the AI breaks a task into steps, proposes actions, waits for user approval, executes approved operations, and displays the resulting changes.

Plan Mode is designed for strategic planning and brainstorming. It can generate structured plans, break problems into actionable steps, provide interactive plan selection, and optionally use web search for research. Ask Mode provides a simpler interface for direct AI-powered questions and answers.

HelloClaw also provides a Telegram interface that brings agent execution, planning, web scraping, and approval workflows to a remote messaging environment. It supports multi-user session management and allows users to interact with different AI capabilities directly through Telegram.

The agent system includes tools for filesystem operations, shell command execution, web scraping, and search. A key part of the architecture is its approval workflow, where proposed agent actions can be reviewed before execution. The system can also generate diffs to make file changes easier to inspect.

The project follows a modular TypeScript architecture with separate modules for AI configuration, terminal UI, agents, planning, Q&A, and Telegram interactions. TypeScript strict mode and Zod runtime validation are used to maintain type safety across agent tools, plans, and integrations.

The AI layer uses the Vercel AI SDK and OpenRouter provider, allowing access to multiple LLM models through a unified interface. Additional integrations include Firecrawl for web scraping, Telegraf for Telegram, and supporting terminal libraries for an interactive CLI experience.