Setoku

Setoku is a self-hosted knowledge server, built on the Model Context Protocol (MCP), that lets AI assistants query a company's databases and other systems while understanding what the data actually means. Rather than just exposing raw tables, it builds and maintains context, metric definitions, entity documentation, and known pitfalls, and improves that context with each interaction; agents can propose updates to the knowledge base, but only a human can approve changes. Database access is read-only and enforced at the engine level, with row caps, query timeouts, and full audit logging. It can connect to PostgreSQL, GitHub, Vercel, Render, Slack, and Mercury, consolidating everything into a local ClickHouse data lake, and it can generate live dashboards from natural-language requests and publish them as shareable links without any separate deployment step. Setoku runs on a self-hosted Ubuntu VPS, the site cites running costs as low as $5-12/month, and works with any MCP-compatible client, including Claude. It's aimed at teams that want AI-assisted data access without sending their data to a third-party SaaS vendor, from startups to sports analysts and academic researchers. Being self-hosted and open source (Apache 2.0), it requires more setup than a hosted BI tool, and detailed pricing beyond the infrastructure cost estimate isn't published.