In a nutshell: Valmis addresses the security challenge of AI agents through a proxy architecture where containerized agents can only make API requests via secure credential management.
A developer has released Valmis, an open-source alternative to Claude Cowork that connects AI agents with over 100 business applications — with security as core design. The system separates agent runtime and API access through a proxy architecture.
The author developed Valmis after experiences with OpenClaw (presumably OpenAI API tools), which was difficult to integrate with APIs and third-party services. The existing tool was not designed for productive collaboration. Valmis addresses this problem through dedicated design for enterprise workflows with over 100 supported applications.
The core architecture relies on a proxy system: the Dockerized agent container cannot make direct API requests; instead, it sends requests to the host system with a credential ID. The host executes the actual request and returns only the JSON response. This keeps the agent container completely air-gapped — internet connectivity can be completely disabled. All credentials remain on the host system, not in the container.
The proxy integration supports Google Workspace (Gmail, Docs, Sheets, Drive), Slack, Notion, HubSpot, Salesforce, Figma, and many other services. An additional feature is a workflow builder for multi-stage automation: workflows can be triggered via Cron jobs, webhooks, or app events and support conditions and loops.
Source: github.com · Published 2 July 2026
Lumi AI News — AI-assisted curation pursuant to Article 50 EU AI Act. Paraphrasing and classification by Lumi News Pipeline v1.7.2.