Second Brain Apps: The Complete Guide for 2026

There is no universal best second brain app. A local Markdown vault, a hosted team workspace, a dedicated reader, and a self-hosted knowledge system solve different problems.

Choose the operating model before comparing feature lists.

The jobs to compare

A useful second brain may need to:

• capture sources, passages, files, and thoughts;
• write notes and preserve explicit links;
• retrieve stored text predictably;
• reveal relationships without hiding provenance;
• support reading and annotation;
• turn evidence into a shared deliverable;
• export data in documented formats;
• fit the amount of infrastructure you want to operate.

No tool is equally strong at every job.

How the tools compare

The numbered order is the order we would suggest trying them in for a general reader who wants help capturing and making sense of what they read and watch. Your own constraints, especially self-hosting and file ownership, can reverse it.

#	Tool	Best for	Where your data lives
1	Recall	AI Second brain app for saving articles, videos, podcasts, and PDFs and getting AI summaries back	Hosted service with local-first apps
2	Arivu	Self-hosted sources and Notes with inspectable links	Self-hosted server
3	Obsidian	Local Markdown writing and plugins	Local files
4	Notion	Shared pages and team databases	Hosted workspace
5	Readwise Reader	Reading and highlight workflows	Hosted reader
6	Raindrop.io	Visual bookmark organization	Hosted bookmark manager
7	wallabag	Open-source read-it-later	Self-hosted server

1. Recall: AI summaries for what you read and watch

Recall is a popular second brain app built around one fast loop: save something, get a summary, and find it again through connections and chat. It captures articles, YouTube videos, podcasts, PDFs, and more through browser extensions for Chrome, Edge, Firefox, and Safari, mobile apps for iOS and Android, and a web app.

What it does well:

• Capture to understanding in one step. Saved items get summaries, automatic tags, and links to related items, so a pile of sources starts to look like a map without manual filing.
• Video and podcasts are supported. Recall’s FAQ lists support for YouTube videos up to 10 hours and PDFs up to 300 pages, and its homepage lists podcasts and TikToks as content types.
• Resurfacing and testing. Augmented browsing shows related saved items while you browse, and quizzes follow a spaced repetition schedule.
• Chat over what you saved. You can ask questions of your own library and save answers to a notebook.
• Open integrations and export. Recall’s documentation covers an MCP server and an API, and its FAQ says you can export all your data as zipped Markdown files from Settings.

The tradeoffs:

• It is a hosted service. Recall describes itself as local-first, with backup and sync stored on Google Cloud in Belgium, according to its FAQ. It offers no self-hosting option, so you rely on the vendor to keep operating.
• Where AI processing happens is not spelled out. Its FAQ and privacy policy state it does not train AI models on your content, but neither says which providers receive your content when AI features run.
• The free plan is limited. At the time we checked, the Free plan on its pricing page allowed 10 AI summaries a month, and automatic tagging, AI chat, and quizzes sat on paid plans.

Choose it when you want the most complete save-summarize-recall experience and do not want to operate infrastructure. Pick another tool when your data must stay on hardware you control.

2. Arivu: self-hosted connected knowledge

Arivu is for people who want bookmarks and Notes in one self-hosted workspace. Home, Library, Notes, and Search are primary; Review and Graph are secondary.

It runs as a single Go binary with SQLite. Capture, full-text Search, notes, links, and graph structure work without AI. Source conversations and quizzes require a configured provider and consent.

It is open source on GitHub. The Features page maps what it does today, and the self-hosting guide covers running it.

Choose it when self-hosting, source-backed thinking, and inspectable relationships matter. Avoid it when you want a managed service with no operational responsibility.

3. Obsidian: local Markdown notes

Obsidian stores notes as Markdown files in a local vault and supports links, backlinks, Graph view, and community plugins. Its official Web Clipper saves pages into the vault. Choose it when the file-based writing environment is the primary product.

Pair it with a source tool only if web capture and reading evidence are separate enough to justify another system.

4. Notion: shared pages and structured workspaces

Notion supports shared pages, databases, a Web Clipper, and workspace exports. Choose it when collaboration, structured team information, and shared deliverables matter more than self-hosting.

It can hold research, but a private source workspace may still help when raw evidence should remain separate from the polished team artifact.

5. Readwise Reader: reading and highlights

Readwise Reader is centered on reading, annotation, and highlight export. Its official export documentation describes Markdown annotation exports, library CSV, OPML, and full content downloads.

Choose it when the reading experience and highlight workflow are the core need.

6. Raindrop.io: hosted bookmark organization

Raindrop.io combines collections, tags, highlights, files, search, and exports in a polished hosted bookmark manager. Pro adds features including full-text content search and web archives, and the help documentation covers how each works.

Choose it when visual organization and cross-device hosted convenience matter more than a broader notes-and-graph system.

7. wallabag: self-hosted read-later

wallabag is a focused read-it-later application that saves readable article content. It supports capture, tags, search, filters, exports, and self-hosted installation, all described in its official documentation.

Choose it when you want open-source read-later without expanding the system into a general second brain.

A fair selection process

Test the same five actions in each candidate:

1. Capture a difficult page and a short note.
2. Find it with an exact phrase a week later.
3. Connect it to another idea.
4. Export the item and its annotations.
5. Explain how you would restore the full system.

The best tool is the one whose strengths align with your daily work and whose tradeoffs you are willing to maintain.