I recently tried using OpenClaw for my agentic workflows. It is an impressive project, but the complexity was a hurdle for my use case. It contains over 400,000 lines of code and requires significant configuration. I found myself managing the infrastructure more than the tasks.

I wanted a minimalist tool that followed the Unix philosophy.

I built Claude-Pipe. It is a 1,000-line bridge that connects Anthropic’s Claude Code CLI to Telegram or Discord.

Technical details:

Inherits existing state. It builds on top of your local Claude Code setup. This means all your previously configured skills, plugins, and sub-agents work out of the box without extra setup.

Small codebase. At roughly 1,000 lines, the entire project is auditable in a few minutes. This is important for security when giving an LLM access to a terminal or private APIs.

VPS deployment. I run this on a private VPS. It provides a persistent, secure environment for sensitive integrations, such as the Gmail-reading skill I use to scan subjects via mobile.

Model flexibility. The CLI allows routing to different models, including third-party options like MiniMax-M2.1, while keeping the same interface.

The goal was to stop reinventing the platform and just create a pipe for the tools that already work.

Link: https://github.com/georgi/claude-pipe

Show HN: Claude-Pipe – A 1k LOC Bridge from Claude Code to Telegram/Discord
github.com