Apache Superset and Preset

Not the coding orchestrator

Do not confuse with Superset (coding agent orchestrator) in KB-AI, an unrelated product from the superset-sh GitHub org. This note is about the Apache BI tool.

What it is

Apache Superset is an open-source, “modern, enterprise-ready” BI web application: no-code chart builder, SQL Lab editor, dashboards, 50+ supported databases (GitHub README, 2026-10-05). Related: Open-source BI comparison.

Maker and history

Began as a hackathon project by Maxime Beauchemin at Airbnb (creator of Apache Airflow); entered the Apache Incubator in 2017 and became a top-level Apache project in 2021 (Wikipedia, secondary source). Preset is the commercial company offering managed Superset; Superset itself remains an independent ASF project. Licence: Apache 2.0.

Products and editions

• Apache Superset: free, self-host (Docker Compose, Docker image, Helm chart for Kubernetes).
• Preset: managed Superset with an Enterprise tier (Preset MCP is Enterprise, per Preset’s announcement); whether a free account option exists is not verified. Pricing: see https://preset.io/pricing (not fetched).

Data connectivity and modelling

SQLAlchemy-based connections to 50+ databases. Lightweight semantic layer: custom dimensions and metrics on datasets; 6.1 adds dataset folders to organise metrics and columns. Jinja templating in SQL.

Deployment

Self-hosted anywhere (Kubernetes-oriented) or Preset cloud.

AI features as of 2026-10-05

• Superset 6.1 (GitHub release 6.1.0 published 2026-05-13; Preset’s release post is dated 2026-06-11, so the post followed the tag by about a month): new opt-in MCP service giving AI assistants structured, authenticated access; JWT auth and field-level access controls; optional fastmcp dependency, runs as a separate process (superset mcp run). Open source, not GA-labelled by me.
• Per Preset repo recaps, the MCP tool catalogue kept growing through June-July 2026, including semantic-layer discovery tools (secondary vendor source).
• Preset MCP: per Preset’s announcement, launched 2026-04-01 for Enterprise customers only; the vendor states 20 tools across charts, dashboards, datasets and SQL Lab, the same RBAC, row-level security and column restrictions as the UI, and use with Claude Desktop and Cursor.
• Preset Chatbot: built-in conversational interface, beta as of the April 2026 announcement.
• A third-party mcp-superset package on PyPI reportedly exists (not vetted).

Strengths and weaknesses (opinion)

• Strengths: permissive Apache 2.0 licence; 50+ supported databases per the README; Preset’s 6.1 release post counts 136 contributors, 69 of them first-time.
• Weaknesses: (opinion) likely heavier to operate and more analyst/engineer-oriented than Metabase; the open-source edition’s AI feature noted above is the MCP service, and no built-in chat was found for it (Preset Chatbot is a Preset feature).

Self-learning

• Preset and Superset Knowledge Base: https://docs.preset.io - Preset, getting started to advanced Jinja templating, free; weekly office hours, workshops (opened 2026-10-05).
• Preset blog release recaps: https://preset.io/blog/apache-superset-6-1-release/ and repo recaps (opened).
• Apache Superset docs: https://superset.apache.org/docs/intro (loads, HTTP 200, 2026-10-07).
• No vault video summaries found in yt-base.

Sources

• https://github.com/apache/superset, https://en.wikipedia.org/wiki/Apache_Superset
• https://preset.io/blog/apache-superset-6-1-release/, https://preset.io/blog/preset-mcp-announcement/, https://docs.preset.io (all 2026-10-05)

Open items

• Preset company funding/history not verified. Preset pricing page loads (https://preset.io/pricing) but tiers not read.
• Re-checked 2026-10-07: https://api.github.com/repos/apache/superset/releases/latest (6.1.0, 2026-05-13, Apache-2.0); Preset 6.1 and MCP posts (June 11 2026 post; April 1 2026 MCP launch, Enterprise only, 20 tools, Claude Desktop/Claude Code/Cursor, Chatbot beta); Wikipedia (secondary: Airbnb hackathon, Incubator 2017, top-level 2021).