mindshub: a unified workspace for delegating multi-step tasks to open-source agents and publishing their output as shareable artifacts

What it solves

MindsHub Cowork is a unified workspace designed to automate multi-step knowledge work—such as research, analysis, and reporting—by delegating tasks to open-source agents. It allows users to connect their data sources and turn agent-generated output into shareable, published artifacts like dashboards and documents.

How it works

The platform acts as a superproject that integrates a desktop/web app, an agent backend, and a data engine. It features a Model Router to switch between frontier models (e.g., Claude, GPT, Gemini) and open models (e.g., DeepSeek, Qwen, Kimi) without needing separate keys for each. It uses interchangeable open-source agent harnesses (such as Anton and Hermes) and a secure data vault to link systems like BigQuery, Postgres, Gmail, and Notion.

Who it’s for

It is built for knowledge workers, including creators, strategists, and operators, who need to automate repetitive reading and writing tasks or build internal AI tools and artifacts without engineering experience.

Highlights

• Connected Data: Securely links to various data systems (BigQuery, Postgres, HubSpot, etc.) without exposing raw keys to agents.
• Model Router: Seamlessly switch between multiple frontier and open-source LLMs.
• Artifacts: Converts agent output into published documents, dashboards, apps, and code via live URLs.
• Flexible Deployment: Supports cloud, VPC, on-prem, air-gapped, and hybrid infrastructure.
• Memory & Scheduling: Includes cross-session memory, a reusable skill library, and the ability to schedule tasks.