ClawScale

ClawScale is an open-source platform for deploying AI chatbots across many messaging platforms without rebuilding webhook infrastructure for every channel.

It connects AI backends such as OpenClaw, Claude, GPT, OpenRouter, self-hosted models, or local CLI agents to messaging surfaces like WhatsApp, Discord, Slack, Telegram, Teams, LINE, Signal, WeChat, and more.

Product Goal

Most teams that want IM chatbots end up gluing together webhook handlers, queues, user-state stores, provider adapters, and AI API calls. ClawScale replaces that repeated integration work with a single gateway and dashboard.

The key design goal was multi-tenant isolation: many users can chat with AI through the same deployed system while their conversations, memory, and state stay separated.

What I Built

• A multi-tenant dashboard for channels, AI backends, conversations, users, settings, and onboarding.
• Messaging adapters for platforms including WhatsApp, WhatsApp Business, Discord, Slack, Telegram, Teams, LINE, Signal, Matrix, WeChat, and WeCom.
• AI backend routing for OpenAI, Anthropic, LangChain, OpenRouter-compatible endpoints, Pulse, OpenClaw, and custom OpenAI-compatible models.
• A CLI bridge package that connects local agents like Claude Code to ClawScale over WebSocket, without needing a public IP.
• End-user onboarding pages that let tenants share a single branded connection URL.
• Dockerized deployment with Prisma-backed persistence and test coverage around auth, routing, validation, slash commands, and backend behavior.

Architecture

ClawScale is a pnpm monorepo with two main packages:

• packages/app/: full-stack app with API routes, web dashboard, Prisma schema, and shared types.
• packages/clawscale-cli-bridge/: CLI/WebSocket bridge for tunneling local AI agents into the hosted gateway.

This separation keeps the gateway layer independent from the agent layer, so teams can scale messaging infrastructure and AI execution separately.