The repository named "mindsdb" no longer hosts the SQL-over-everything engine many remember; in 2026 the project rebranded — mindsdb.com now 301-redirects to mindshub.ai — and repurposed itself entirely. What lives here today is the Minds-Cowork Platform: a general-purpose AI workspace aimed not at database engineers but at knowledge workers — creators, strategists, and operators — who want AI agents they can actually control and run anywhere.
The core idea is delegation over chat. Instead of a chat box that answers questions, you describe a task and get finished, shareable work back — documents, dashboards, reports, small apps — produced by agents wired to your own data.
What Sets It Apart
- Agent and model layers are interchangeable: swap between open agent harnesses (Anton, Hermes, OpenClaw, NanoClaw) and LLM providers (Anthropic, OpenAI, Google, DeepSeek, Qwen, Kimi) from a dropdown, and your built work, memory, and skills carry over — no rewrite, no lock-in.
- Data connects through a secure vault (BigQuery, Postgres, Gmail, Drive, Notion, Linear, and more) so agents act on real sources without ever seeing raw credentials.
- Outputs are live artifacts, not transcripts: tasks yield usable docs, dashboards, or code you can publish to a shareable URL.
- It runs where your data is — cloud, VPC, on-prem, air-gapped, or hybrid — as a web app or macOS/Windows desktop build.
Who It's For
Great fit if you want a controllable, deployable agent workspace that turns plain-language requests into finished artifacts on top of your own systems, and you care about avoiding agent or model lock-in. Look elsewhere if you came for the legacy MindsDB — in-database ML and federated SQL queries are gone from this codebase — or if you just need a conversational chatbot rather than task delegation that produces deliverables. As a freshly repositioned platform, expect the surface and docs to keep shifting.
