Show HN: I built an open-source alternative to Claude Cowork

Valmis is an open-source "AI coworker" platform — basically a more enterprise-friendly alternative to tools like OpenClaw (an open-source AI agent project) and Claude Cowork (Anthropic's official agentic coding/productivity tool). The key problem it solves: AI agents that interact with third-party APIs (e.g., Google Docs, Slack, Salesforce) pose a security risk because the agent's code directly handles credentials. Valmis uses a "proxy architecture" — the AI agent runs inside a Docker container that has no direct internet access; instead, it asks the host machine to make API calls on its behalf, passing only a credential ID. This means even if the agent is compromised or malicious, it never touches your actual API keys or tokens. It also supports automated multi-step workflows (triggered by schedule, webhooks, or app events) with conditions and loops, positioning it as a competitor to both AI agent frameworks (like AutoGPT, CrewAI) and low-code automation tools (like Zapier or Make).