Heptabase

Heptabase is a visual whiteboard-first knowledge base. Instead of pages in a folder tree or blocks in a database, the core unit is a card note that lives on an infinite whiteboard, and whiteboards can be nested inside other whiteboards. That nesting is the product’s signature move: a research project can hold a whiteboard for sources, another for arguments, another for an outline, each one a zoomable layer inside the parent canvas.

Card notes carry PDF highlights, YouTube transcripts, and web clips onto the whiteboard directly, so source material and synthesis sit in the same visual space rather than in separate apps. The newer AI Tutor, built on Anthropic’s Claude Agent SDK and launched March 31, 2026, takes a learning goal, generates a full course syllabus, runs interactive lessons with checkpoints, and converts completed lessons into structured whiteboard notes on request.

Heptabase is built by Heptabase Labs, a Y Combinator W22 company founded in 2021 by Alan Chan, Yeefun Lin, and Ethan Wu, backed by Y Combinator, Kleiner Perkins, HOF Capital, and Moving Capital. There is no permanent free plan. New users get a trial of up to 7 days or until trial AI credits run out, whichever comes first, then a paid subscription is required to continue.

What it actually is

Heptabase organizes knowledge around two ideas: card notes and whiteboards. A card note is a single, atomic piece of content, similar to a Notion page or an Obsidian file. A whiteboard is an infinite canvas where you arrange cards spatially, draw connections between them, and group related cards into sections. The differentiator is that a whiteboard can contain another whiteboard as a nested object, so a large research topic can be broken into layered sub-canvases without leaving the visual metaphor.

PDF highlights and YouTube transcript snippets can be dragged directly onto a whiteboard as cards, keeping source material and your synthesis of it in the same visual space. Readwise and Zotero integrations pull highlights and citations in automatically for research-heavy workflows.

The AI Tutor, introduced in 2026, is a separate but integrated feature: tell it a learning goal, it drafts a full syllabus with individual lessons, walks you through them with checkpoints and real-time Q&A, and on request turns completed lesson content into organized whiteboard notes. It runs on Anthropic’s Claude Agent SDK and consumes the same AI credit pool as chat and tagging features.

There is no permanent free plan. Every account starts on a trial capped at 7 days or trial AI credit exhaustion, after which a Pro, Premium, or Premium+ subscription is required.

When to pick Heptabase

• Spatial thinkers. If your mental model for a project is “place things, arrange them, connect them,” the whiteboard-first design fits better than a database or folder tree.
• Nested, multi-layered research. Whiteboards inside whiteboards let a single big topic hold source boards, argument boards, and outline boards as zoomable layers.
• PDF- and video-heavy study. Highlights and transcripts drop onto the canvas as cards, so source review and synthesis happen in the same view.
• Guided AI learning. AI Tutor’s syllabus generation and checkpoint lessons are a genuinely different feature from generic AI chat-with-your-notes.
• Fast onboarding. Reviewers consistently describe Heptabase as more intuitive to pick up than Tana’s Supertag system, often productive within a day or two.

When to pick something else

• You want a free tier first: Obsidian is free for personal use with a large plugin ecosystem, including whiteboard-style Canvas.
• Local-first markdown ownership: Obsidian or Logseq store notes as plain markdown files you control.
• Team workspace and docs: Notion AI is the stronger pick for shared wikis, databases, and company-wide documentation.
• Structured typed-object knowledge: Capacities models people, books, and projects as objects with properties rather than freeform cards.
• Database-style querying over notes: Tana’s Supertag model rewards users willing to invest weeks learning its query system.
• Encrypted, minimal linked notes: Reflect is a simpler, cheaper backlinked-notes app for solo users.

Pricing

Heptabase pricing verified 2026-07-02 via heptabase.com/pricing:

There is no free plan. New users get a trial of up to 7 days or until trial AI credits (set at Premium-level, 1,800) are used, whichever comes first. Annual billing saves 25% versus month-to-month across all three tiers.

Note that Heptabase’s own support-article pricing FAQ lists higher monthly rates than the live marketing pricing page in places (for example Pro at $11.99/month monthly-billed versus the $8.99/month annual rate), and some early-bird subscribers from before the 1.0 launch in 2023 retain older locked-in pricing indefinitely. Always confirm the number shown at checkout.

Against the alternatives