Heptabase Community Discussions

WhatAI Heptabase Forum Discussions

Heptabase is a visual knowledge and research workspace for people who need to understand complex topics, not merely collect notes. It combines note cards, whiteboards, tags, structured tag databases, journals, tasks, highlights, PDFs, web cards, media transcripts, mind maps, collaboration, and cross-device applications in one connected system. Cards can be reused across several whiteboards, while sections, links, colors, spatial placement, and nested boards help users show relationships without duplicating the underlying note. This makes Heptabase especially relevant to students, researchers, writers, founders, educators, analysts, and lifelong learners who build understanding across many sources. The core workflow starts with capture, reading, or import. Users can create notes, record voice notes, clip pages, upload files, import Markdown, sync Readwise highlights, connect Zotero collections, transcribe media, or work through daily journals. They can then turn useful material into cards, place those cards on whiteboards, tag them, add properties, connect related ideas, and revisit the resulting structure through search, backlinks, tag databases, or visual navigation. Heptabase supports desktop apps for macOS, Windows, and Linux, mobile apps for iOS and Android, and a web app. Desktop data is held in a local database, while cloud sync is enabled by default for normal multi-device use. This is different from an ordinary folder of user-managed Markdown files. Heptabase can export notes as Markdown and supports manual full backups, but whiteboard layout, databases, metadata, attachments, and application behavior should be tested during any migration. Automatic cloud backups do not include attachments, so a separate manual full backup remains important. Heptabase now has a substantial native AI layer. AI Agent can search and read selected content, cards, whiteboards, tag databases, PDFs, images, transcripts, and other workspace context, then answer questions or edit supported knowledge-base objects. AI Tutor can build guided learning materials from sources and let users review changes. AI actions can summarize, translate, transform, or create mind maps from cards. Space Instructions and reusable skills can shape how the agent works. Users can pay through Heptabase credits, enable on-demand usage, connect their own OpenAI or Anthropic API keys, use supported OpenRouter models, or connect an eligible ChatGPT plan for supported routes. The composer identifies which billing route will fund a request. These routes are not interchangeable. Plan credits reset and do not roll over, on-demand charges can exceed a selected cap slightly when a final request completes, personal API keys are billed and governed by the external provider, and an eligible ChatGPT subscription is separate from the Heptabase subscription. Premium and Premium+ continue to offer basic read, search, and answer chat after included credits are exhausted, but that mode cannot create or edit content. Heptabase also exposes two agent integration routes. Its OAuth-based MCP server lets compatible assistants search cloud-synced knowledge, inspect whiteboards and objects, read journals and PDF pages, then create cards or append to journals through supported tools. The desktop CLI lets compatible coding agents work through the running local app, including cards, journals, search, tags, whiteboards, properties, parsed PDF text, transcripts, and AI Tutor content. Heptabase explicitly warns against modifying its local database through unofficial methods. Security and privacy require careful reading. Heptabase says cloud data is stored in AWS us-west-1, with AES-256 server-side encryption at rest and SSL/TLS in transit. This is not described as end-to-end encryption. Its current terms say normal integrated AI services can send content, including text, audio, and video, to OpenAI, Google, or Anthropic, while some AI suggestions and PDF parsing use self-hosted models on Runpod infrastructure. The company says it does not train its AI models on user content and has provider agreements preventing general foundation-model training, but providers may retain data temporarily for abuse monitoring. Those negotiated protections do not apply to personal API keys. The terms also prohibit using AI services to process sensitive personal data. Space search is on by default in AI Agent conversations. The system can search across the enabled Space and normally sends a selected subset of relevant items to the model. Users cannot exclude one sensitive card while leaving the rest of that Space searchable, so sensitive workflows should disable Space search and add only approved context manually. Pricing is subscription based. Standard Pro costs $11.99 monthly or $107.88 yearly, equivalent to $8.99 per month, and includes the core knowledge workspace plus 100 monthly AI credits and Gemini-series models. Premium costs $23.99 monthly or $215.88 yearly, equivalent to $17.99 per month, and adds 1,800 credits, premium models from OpenAI, Google, and Anthropic, AI Tutor, unlimited PDF uploads and parsing, and basic chat after the credit limit. Premium+ costs $71.99 monthly or $647.88 yearly, equivalent to $53.99 per month, and raises the allowance to 8,100 credits with otherwise similar Premium features. New buyers receive a seven-day trial or the Premium credit allowance, whichever is exhausted first. Heptabase has no directly accessible full free personal plan, although invited non-subscribers can use a restricted collaborator plan for shared whiteboards. Refunds are handled case by case in the Help Center, while the governing terms describe post-September 2023 fees as non-refundable, so buyers should not treat a refund as guaranteed. Heptabase is strongest when visual structure, source reading, long-term learning, and native AI need to live in the same workspace. It is less suitable for people who want a free personal notes app, plain files as the live source of truth, end-to-end encrypted cloud collaboration, mature enterprise administration, unrestricted AI at one fixed price, or a conventional public REST API. The best trial uses a real research or learning project, measures whether the whiteboards improve retrieval and understanding, tests export and restoration, and records the total AI cost across the actual models and billing routes used.

Popular Discussions

The most-engaged Heptabase conversations in the WhatAI community.

Recent Discussions

The latest Heptabase conversations from the community.

Related Productivity Tools

Blaze AI →Cleanup.pictures →Designrr →Fireflies.ai →

Keep Exploring

Recommended Articles

Best AI Guides

AI Frontiers Conversations

Have a question about Heptabase?

Share your experience, ask for help, or discuss tips with the community. All posts are reviewed by our moderation team.