dalao.ai

Heptabase

Visual note-taking for deep thinking

About Heptabase

Heptabase is a visual knowledge base built on whiteboards and cards for students, researchers and lifelong learners, made by Hepta Platforms, Inc. (founded September 2021, Y Combinator Winter 2022). It is closed source, runs on desktop, web and mobile, with AI chat, an AI tutor, PDF parsing, plus MCP and a CLI for outside AI agents. There is no self-serve free tier; new users get a 7-day trial. Heptabase pricing page (as of 2026-10-25): Pro is $8.99/month billed annually, $11.99/month billed monthly.

EditorialUpdated 12 min read

The Heptabase interface

What it is

Heptabase is a visual knowledge base. Its homepage describes it as an intelligent, visual knowledge base built for students, researchers and lifelong learners, under the tagline "Master anything you learn. Do your best research with AI." The basic unit is the card: each card is a note that can be edited on its own, and cards can be placed on one or more whiteboards and arranged with arrows, sections and mind maps.

Around cards and whiteboards, the homepage lists a block-based editor, bi-directional links, PDF annotation, daily journals, a web clipper, Readwise integration and Zotero integration. The Heptabase 1.0 announcement positions it as a tool for learning and researching complex topics, and its examples of users include product managers, founders, engineers, designers, researchers, graduate students, authors, screenwriters and teachers. In the 2026 changelog, a mobile update on 2026-09-28 added Kanban and table views for tag databases.

It runs on Mac (Intel and Apple Silicon), Windows, Linux, the web, iPhone, iPad and Android. Desktop installers are published through the heptameta/project-meta repository on GitHub, but that repository holds no source code: the README is a single title line, and the GitHub API shows no declared license. The product itself is closed source.

The developer is Hepta Platforms, Inc.; the terms of service, the App Store and Google Play all name this company. The Y Combinator company page lists Heptabase in the Winter 2022 batch, gives 2021 as the founding year and 10 as the team size, and names its co-founders: Alan Chan, Yeefun Lin and Ethan Wu.

Create a world where anyone can effectively establish a deep understanding of anything.

Heptabase Public Wiki homepage

Core capabilities

First, cards on whiteboards. The Heptabase 1.0 announcement explains that users can break long cards into small single-concept cards, organize the structure between topics with nested whiteboards, group cards with sections, adjust their understanding of a topic with auto-layout mind maps, and show their thinking with arrows of different styles. The homepage says searching over 10,000 notes takes less than a second and all notes and files are accessible without an internet connection.

Second, handling reading material. The AI documentation says that when researching a topic users can upload PDFs, YouTube links or playlists, .docx files, plain-text documents and images, and the system puts all of them on a new whiteboard. The help center's plans page states that PDF parsing has no cumulative monthly limit on any of the three paid plans; each PDF must be no larger than 350MB and no longer than 3,000 pages, and parsing does not use AI credits.

Third, AI. Users can add whiteboards, cards, PDFs, videos and journals to an AI chat as context, then drag useful answers back onto the whiteboard. The homepage also has an AI Tutor entry; a changelog entry on 2026-03-31 lets the AI Tutor run on on-demand usage once credits run out. v1.101.1, released on the GitHub releases page on 2026-07-28, lets users use ChatGPT Codex inside Heptabase, and the chat composer shows whether the next message will use the user's ChatGPT plan, the user's own API key or Heptabase AI credits.

Fourth, outside interfaces. A changelog entry on 2026-04-22 introduced the Heptabase CLI; around the same time, the heptameta/heptabase-cli-skills repository began offering skill packs for Claude Code, Codex CLI and OpenCode, under the MIT license. The AI documentation on the public wiki also covers MCP, through which outside AI services can read, search and write to the knowledge base. According to the changelog, the ChatGPT Plugin shipped on 2026-09-23 and the Claude Plugin on 2026-10-09.

Fifth, collaboration. The homepage says friends and teammates can be invited to collaborate in real time for free. The help center explains that new users invited by a paid user join the free plan automatically and can edit whiteboards and cards in spaces where they have collaborator permissions, but cannot use the journal, the personal whiteboard and card library, the tag database, tasks, highlights or AI features.

Who it's for / Not for

Good fit: people who read many PDFs, papers, videos and web pages and need to turn them into a structure for a topic. The Heptabase 1.0 announcement gives examples such as researchers reading papers, graduate students building research frameworks, authors organizing material and product managers researching niche markets; in each case the core move is to put scattered material on a whiteboard, work out the relationships and produce an outcome.

Good fit: visual thinkers who remember information by where it sits in space. Whiteboards can be nested, the desktop app can export a whiteboard as an image, and a mobile update on 2026-09-23 added a whiteboard minimap for navigating on a phone.

Good fit: people already using Claude Code, Codex or ChatGPT who want an AI agent to read and write their personal notes directly. The CLI, MCP, the ChatGPT Plugin and the Claude Plugin are all available, and the changelog says the CLI can read, inspect and modify whiteboards, including their hierarchy, layout, objects, sections, connections and mind maps.

Not a fit: individual users with a zero budget. The help center states that the free plan is available only through a collaboration invite from a paid user and cannot be registered directly; after a subscription ends, everything can still be viewed and exported, but notes cannot be created or edited.

Not a fit: teams that require open source or self-hosting, or need to audit source code. The product is closed source and the GitHub release repository holds no source code; the terms of service state that features such as AI chat, transcripts, voice notes and insight generation send content to OpenAI, Google and Anthropic for processing.

Not a fit: anyone under 18. The terms of service require users to be at least 18, and people under 18 may not access or use the service.

How it compares

Versus Obsidian: the Obsidian pricing page (as of 2026-10-25) states that the app is free with no sign-up required; Sync is $4 per user per month billed annually or $5 per user per month billed monthly, and Publish is $8 per site per month billed annually or $10 per site per month billed monthly. Heptabase has no free tier open to direct sign-up, and its cheapest plan, Pro billed annually, works out to $8.99 a month, with whiteboards, PDF annotation, AI and sync in one subscription.

Versus Logseq: Logseq is open-source note-taking software; the GitHub API shows the AGPL-3.0 license and, as of 2026-10-25, 45,369 stars, and the code can be used and modified freely under that license. Heptabase is closed source and sold by subscription.

Versus Scrintal: the Scrintal pricing page (as of 2026-10-25) lists Pro at $9 a month billed annually; Scrintal likewise focuses on visual thinking with documents and boards, and supports real-time collaboration. Heptabase Pro billed annually works out to $8.99 a month.

Heptabase Pro includes 100 AI credits a month, plus the CLI, MCP and the ChatGPT and Claude plugins; the Pro plan on the Scrintal pricing page lists unlimited docs and boards, unlimited file uploads, real-time collaboration and priority support.

Versus doing research in a general AI chat app: the Heptabase AI documentation points out that complex topics involve connections between many concepts and that linear chat records are hard to review; its approach is to drag AI answers onto a whiteboard and organize them alongside existing notes.

Getting started

Step one is downloading. The download page offers Mac (Intel and Apple Silicon), Windows, Linux, iPhone, iPad and Android versions plus a web clipper. Desktop installers come from the heptameta/project-meta releases page on GitHub; version updates follow the official changelog.

Step two is the trial. The homepage offers a 7-day trial; the help center explains that free-plan users can also start a 7-day free trial by clicking the Upgrade button on the right side of the screen, and that accounts created before 2025-02-07 cannot join the free plan but can start a trial and keep collaborating in the inviter's space after the trial ends.

Step three is bringing in material. A new topic whiteboard can hold PDFs, YouTube links and existing notes, and the research-a-topic button in the left sidebar uploads several sources at once; a changelog entry on 2026-02-23 added a Zotero integration (Beta).

Step four is learning three basic moves. The public wiki's introductory videos come in three parts: sense-making with whiteboards, organizing topics with nested whiteboards and tab groups, and managing card databases with tags and properties.

Step five is connecting an AI agent. Claude Code or Codex read and write notes through the Heptabase CLI and the skill packs in heptameta/heptabase-cli-skills, whose README covers installation; ChatGPT and Claude have matching plugins, and MCP is also available.

Pricing & license

The Heptabase pricing page (as of 2026-10-25) has three plans and shows annual billing: Pro at $8.99 a month, Premium at $17.99 a month and Premium+ at $53.99 a month, with annual billing marked "Save 25%". Switched to monthly billing, the same page shows $11.99, $23.99 and $71.99 a month.

The help center's plans page lists the yearly totals for annual billing: Pro $107.88 a year, Premium $215.88 a year and Premium+ $647.88 a year. Early-bird Pro costs $9.99 a month billed monthly or $83.88 a year billed annually, and early-bird users keep that price permanently; Premium and Premium+ have no early-bird price. Prices are in USD, and VAT or other taxes may apply.

The plans differ mainly in AI. Pro includes 100 AI credits a month and its AI agent uses Gemini models only; Premium includes 1,800 credits a month, supports Gemini, OpenAI and Anthropic models, and adds the AI Tutor and unlimited PDF uploads; Premium+ includes 8,100 credits a month plus a 33% discount on AI credits. AI credits cover the AI agent, the AI Tutor, video transcripts, insight generation and voice notes.

The free plan is available only through a collaboration invite and is limited to editing whiteboards and cards in spaces where the user has permission. The help center states that there is currently no educational, nonprofit or NGO discount.

On refunds, the help center says that if a user was charged recently and has not used Heptabase or its AI features during the current billing period, the most recent charge can generally be refunded in full, and other cases are reviewed individually. The terms of service, by contrast, state that purchases made before 2023-09-01 could be refunded within 7 days, while for purchases on or after 2023-09-01 payment obligations are non-cancelable and fees paid are non-refundable.

License: the product is closed-source proprietary software, and the release repository heptameta/project-meta declares no license. In the heptameta organization, the heptabase-cli-skills and heptabase-agent-plugins repositories are MIT, and heptabase-explainer-studio is Apache-2.0.

Risks & limits

Where data goes. Section 9 of the terms of service states that AI chat, transcripts, voice notes and the insight generator send text, audio and video content to OpenAI, Google and Anthropic; AI suggestions and PDF parsing are processed with self-hosted models on infrastructure provided by Runpod. Third-party AI providers may retain data for a limited period, for example 30 days, for abuse monitoring, and immediate deletion requests may not be processed during that time.

The terms also state that Hepta Platforms does not use user content to train its own models and has agreements with third-party AI providers that they will not use user data to train their general foundation models; users retain all rights to their input and own the AI-generated output. When users bring their own API key, data handling is governed by their agreement with that provider.

Privacy policy date. Its last update is dated 2021-12-17, it lists Amplitude and Firebase as analytics services, and it does not mention AI providers; AI-related data handling sits in the terms of service, last updated 2025-12-01.

The refund rules do not match. The terms of service say subscriptions purchased on or after 2023-09-01 are non-refundable, while the help center's refund article says unused periods can generally be refunded in full. Where they conflict, the text of the terms and the answer from support govern.

Export. The help center's data-portability article says all notes can be exported as Markdown from the backup settings on desktop or web, but the receiving app may not preserve Heptabase's whiteboard layouts, relationships or metadata in the same way. After a subscription ends, notes can still be viewed and exported, but not edited.

The mobile descriptions differ in age. The App Store and Google Play descriptions say the mobile app gives full access to all cards and journals and lets users read all their whiteboards. The 2026 changelog, meanwhile, records a smoother mobile whiteboard toolbar, a whiteboard minimap, finding content within a whiteboard and asking AI about whiteboards. As of 2026-10-25, the App Store shows 71 ratings averaging 4.3; Google Play shows 225 reviews averaging 3.8 and more than 10,000 downloads.

Team size. The Heptabase 1.0 announcement says the team grew from 3 to 6 people over two years; the Y Combinator company page gives a team size of 10.

How it got here

It began with a vision essay. "The Context", the first essay in the vision series on the public wiki, was published on 2020-06-29 by co-founder Alan Chan, two years before work on the early alpha began. The GitHub API shows the heptameta organization was created at 2020-12-11 19:41 (UTC+8).

The original name was Meta. The Star Rocket podcast's episode 128 page, published on 2021-10-06, recalls that episode 127 hosted Alan Chan, founder of Project Meta, to talk about the new note-taking app Meta the team was building; an update dated 2021-11-17 on the page says Meta had been renamed Heptabase. The release repository heptameta/project-meta was created at 2021-10-11 21:48 (UTC+8) and keeps the old name.

The company was founded in September 2021. The help center's data-portability article says Heptabase raised nearly $1.7 million in seed funding in early 2022 from investors including Y Combinator, HOF Capital, Kleiner Perkins and Moving Capital. The same article says the company has been cash-flow positive since its first year, has not spent any investor money over the past four years, and gets 90% of its growth organically.

The Heptabase 1.0 announcement on the official wiki (https://wiki.heptabase.com/version-one) announces the official launch of 1.0; the official 2023 changelog page dates that update 2023-08-31 (https://wiki.heptabase.com/changelog/2023). The announcement says the product had gone through 8 months of closed alpha and 18 months of public beta, with more than 300 versions, and had been an offline-first desktop app since the closed alpha. It adds that real-time sync across devices launched the previous October, a full-featured web app launched that January, and the mobile apps launched that August. At the time, export supported Markdown only.

The November 2024 AMA set out three directions for the next stage: collaboration and sharing, AI-augmented learning and research, and usability improvements. The answers note that collaboration work started that April, and that making Heptabase collaborative and offline-first at the same time raised a large number of technical problems.

The 2026 changelog page lists desktop updates from 2026-01-05 and mobile updates from 2026-01-06; both desktop and mobile apps are updated continuously. The public roadmap, updated 2026-10-22, lists work in progress on whiteboard and mobile performance, faster startup on web, mobile and desktop, better AI agent integration with Codex and Claude Code, and interactive cards; an AI agent LLM Wiki is done and in internal testing but not yet shipped.

As of 2026-10-25, the earliest stable release still on the GitHub releases page is v1.98.6 at 2026-07-16 05:46 (UTC+8); there is also a prerelease, v1.29.0-alpha.0, published at 2024-02-28 14:16 (UTC+8). Version updates follow the official changelog.

Our verdict

Heptabase organizes reading material with whiteboards and cards, with built-in PDF parsing, AI chat and an AI Tutor; in 2026 it added a CLI, MCP and ChatGPT and Claude plugins.

Its limits: it is closed source, has no free tier open to direct sign-up, and when notes are exported as Markdown the receiving app may not preserve whiteboard layouts, relationships or metadata.

Why it stands out

Heptabase is a thinking tool built around a whiteboard of cards. It is designed for the moment when you have read a great deal and need to make sense of it, and it takes information density as a design problem rather than a constraint.

Alan Chan, Yeefun Lin and Ethan Wu co-founded it and went through Y Combinator, and the application balances visual calm with the ability to hold hundreds of notes in view. Cards export in full to Markdown, whiteboards as images and tag databases to CSV or Excel, and it works offline.

It organises notes as cards on whiteboards, works offline, and exports cards to Markdown, whiteboards as images and tag databases to CSV or Excel.

Sources

Official

  1. Heptabaseheptabase.com · Accessed Oct 25, 2026
  2. Download | Heptabaseheptabase.com · Accessed Oct 25, 2026
  3. Pricing | Heptabaseheptabase.com · Accessed Oct 25, 2026
  4. Plans and pricing | Heptabase Help Centersupport.heptabase.com · Accessed Oct 25, 2026
  5. Does Heptabase offer a free plan? | Heptabase Help Centersupport.heptabase.com · Accessed Oct 25, 2026
  6. What happens to my notes after my subscription or free trial ends? | Heptabase Help Centersupport.heptabase.com · Accessed Oct 25, 2026
  7. Business sustainability and data portability | Heptabase Help Centersupport.heptabase.com · Accessed Oct 25, 2026
  8. How do I request a refund? | Heptabase Help Centersupport.heptabase.com · Accessed Oct 25, 2026

Updates & reading

No articles or updates yet. Claim it and connect a feed.