State of Knowledge Management & Knowledge Repository Building (April 2026)

Author: kfchou  |  GitHub stars: 94  |  Licence: MIT

The most complete implementation of Karpathy’s LLM Wiki pattern as a Claude Code plugin. Provides five commands that cover the full lifecycle of a knowledge base: bootstrapping, ingestion, querying, linting, and updating.

Key features:

• wiki-init — bootstrap a new wiki with directory structure and schema
• wiki-ingest — add sources with automatic backlink auditing
• wiki-query — question the wiki, optionally saving answers as new pages
• wiki-lint — health audits for contradictions, orphan pages, broken links, and knowledge gaps
• wiki-update — revise pages with diffs and citations

Install:

/plugin marketplace add kfchou/wiki-skills
/plugin install wiki-skills@kfchou/wiki-skills

Author: toolboxmd  |  GitHub stars: 67  |  Licence: unspecified

Two complementary Claude Code skills for building persistent knowledge bases: a general-purpose knowledge base skill and a project documentation skill. Designed as a minimal, drop-in implementation of the Karpathy pattern.

Key differentiator: Ships as two separate SKILL.md files — one for general knowledge (research, learning, reference material) and one for project-specific documentation (architecture decisions, API contracts, deployment notes). Auto-triggers on source additions, synthesis questions, and health checks.

Architecture: Three-layer model (raw sources, wiki, schema) with three core operations (ingest, query, lint).

Install:

cp -r karpathy-wiki ~/.claude/skills/wiki
cp -r karpathy-project-wiki ~/.claude/skills/project-wiki

Author: Pratiyush  |  GitHub stars: 160  |  Licence: MIT

The most ambitious LLM Wiki implementation. Transforms session transcripts from Claude Code, Codex CLI, Cursor, Gemini CLI, and Obsidian into a searchable, interlinked knowledge base. Goes beyond the basic pattern by adding static site generation, AI-consumable exports, and an MCP server.

Key features:

• Converts .jsonl session files to organised Markdown pages
• Static site generation with search, dark mode, and keyboard navigation
• AI-consumable exports (llms.txt, JSON-LD knowledge graphs)
• Confidence scoring and 5-state lifecycle management for knowledge items
• 14 linting rules for knowledge quality
• Obsidian vault integration with Dataview queries
• MCP server with 12 production tools for agent access
• Zero npm dependencies

Install: git clone + ./setup.sh or pip install -e .

Author: Ar9av  |  Licence: unspecified

Framework for AI agents to build and maintain an Obsidian-compatible wiki using the Karpathy LLM Wiki pattern. Bridges the gap between Claude Code’s file-based workflow and Obsidian’s rich ecosystem of plugins, graph views, and community themes.

Key differentiator: Outputs are native Obsidian vault files — open the wiki directory in Obsidian and get graph views, backlinks, search, and 1,800+ community plugins for free.

Author: cyanheads  |  GitHub stars: 467  |  Licence: Apache 2.0

The leading MCP server for Obsidian vault access. Enables any MCP-compatible AI agent (Claude Code, Codex, Cursor, etc.) to read, write, search, and manage notes in an Obsidian vault through the Local REST API plugin.

Tools provided:

• obsidian_read_note, obsidian_update_note — full CRUD on notes
• obsidian_search_replace — find-and-replace across vault
• obsidian_global_search — vault-wide search with regex support
• obsidian_list_notes — directory listing and navigation
• obsidian_manage_frontmatter, obsidian_manage_tags — metadata operations

Transport: stdio and HTTP. In-memory caching for performance.

Install: npx obsidian-mcp-server (requires Obsidian Local REST API plugin)

Author: Notion (makenotion)  |  Licence: open source

The official Notion MCP server — the most polished cloud-knowledge integration available. Enables AI agents to search across an entire Notion workspace, create documentation, manage databases, and retrieve context from any page.

Key features: Converts all Notion content to Markdown optimised for AI agent consumption. OAuth authentication. Full workspace search, page creation, and database management.

Install (Claude Code):

claude mcp add --transport http notion https://mcp.notion.com/mcp

Author: khoj-ai  |  GitHub stars: 28,000+  |  Licence: AGPL-3.0

Open-source “AI second brain” — the leading self-hostable AI knowledge assistant. Indexes your documents and answers questions with citations, combining personal knowledge with web search.

Key features:

• Chat with any LLM backend (llama3, Qwen, Gemma, Mistral, GPT, Claude, Gemini, DeepSeek)
• Indexes images, PDFs, Markdown, Org-mode, Word, and Notion files
• Accessible from browser, Obsidian plugin, Emacs, Desktop, Phone, and WhatsApp
• Build custom agents and schedule automations
• Deep research mode for complex multi-step queries
• Scales from on-device to cloud-scale enterprise

Install: pip install khoj or self-host via Docker. Cloud app at app.khoj.dev.

Author: Mem Labs  |  Licence: Proprietary (freemium)

AI-powered note-taking that eliminates manual organisation entirely. The thesis: if AI can understand your notes well enough, you never need folders, tags, or manual structure.

Key features: AI automatically organises, tags, and surfaces notes based on context. “Chat” with your entire knowledge base. Synthesises answers from your notes. No folders or manual structure needed.

Key differentiator: Zero-organisation philosophy — the opposite of Zettelkasten. You dump everything in, AI handles retrieval and connection.

Author: Glean Technologies  |  Licence: Proprietary (enterprise)

Enterprise AI-powered knowledge management. Connects to all enterprise apps (Google Workspace, Slack, Confluence, Jira, Salesforce, etc.) and provides a unified AI search and assistant across the entire organisation’s knowledge.

Key differentiator: Enterprise-grade with SSO, compliance, and admin controls. Auto-generates knowledge articles from scattered information. The leading “enterprise brain” platform.

Author: Dynalist Inc.  |  Users: 5 million+  |  Licence: Proprietary (free for personal use)

The dominant Markdown-based personal knowledge base. Plain Markdown files with bidirectional linking, graph visualisation, and an ecosystem of 1,800+ community plugins. The de facto standard for Zettelkasten-style PKM.

Key features: Local-first plain Markdown files; bidirectional [[wikilinks]]; interactive graph view; Canvas for visual thinking; native mobile apps; 1,800+ community plugins; handles 10,000+ notes smoothly.

AI integrations: 45+ MCP servers on PulseMCP; Khoj plugin; Smart Connections plugin; Copilot plugin; native AI features in development.

Pricing: Free for personal use. Commercial $50/user/year. Sync $5/mo (optional).

Author: Logseq Inc.  |  Licence: AGPL-3.0

Open-source outliner for knowledge bases with block-based organisation. Where Obsidian is document-centric, Logseq is block-centric — every bullet point is an addressable, linkable unit.

Key features: Outliner-centric interface; block-level referencing and embedding; daily journal workflow as default entry point; bidirectional linking; completely free and open source.

Key differentiator: Block references enable fine-grained linking — link to a specific bullet point rather than an entire page. Daily journals provide a low-friction capture method. DB version in beta (early 2026) for improved performance.

Author: D&T (formerly siyuan-note)  |  GitHub stars: 42,436  |  Licence: AGPL-3.0

Privacy-first, self-hosted, fully open-source personal knowledge management. Built with TypeScript and Golang. The leading Chinese-origin PKM tool with a growing global user base.

Key features: Fine-grained block-level referencing and embedding; Markdown WYSIWYG with real-time rendering; content blocks as first-class citizens; self-hostable; S3/WebDAV sync; encrypted cloud backup.

Key differentiator: Combines the best of Notion’s block model with Obsidian’s local-first philosophy. 42k+ GitHub stars make it one of the most popular open-source PKM tools globally.

Author: TriliumNext community  |  Licence: AGPL-3.0

Open-source hierarchical note-taking application focused on building large personal knowledge bases. Community-maintained fork since 2024 after the original Trilium project was archived.

Key features: Hierarchical tree structure (notes can appear in multiple locations via cloning); WYSIWYG editor; knowledge graph visualisation; JavaScript-based scripting and custom widgets; per-note encryption; self-hosting via Node.js.

Key differentiator: The scripting system — notes can contain executable JavaScript that automates knowledge management tasks. Cloning (a note appearing in multiple tree locations without duplication) is unique among PKM tools.

Author: Dendron  |  Licence: AGPL-3.0

Hierarchical note-taking tool built as a VS Code extension. Uses dot-namespaced hierarchies (project.backend.auth.oauth) for predictable, scalable knowledge organisation.

Key differentiator: Hierarchy-first approach — while Zettelkasten tools emphasise flat notes with links, Dendron embraces explicit tree structure via naming conventions. Git-friendly files. Predictable paths make it excellent for developer documentation.

Author: Foam community  |  Licence: MIT

VS Code extension for personal knowledge management inspired by Roam Research. Lightweight alternative to Obsidian for developers who prefer to stay in their editor.

Key differentiator: Lives inside VS Code — no separate app needed. GitHub-friendly (designed to be pushed as a repo). Covers most core PKM features (bidirectional links, graph view, daily notes) with minimal overhead.

Author: Notion Labs  |  Licence: Proprietary

The all-in-one workspace for notes, tasks, wikis, and databases. While not local-first, Notion remains the most widely used knowledge management platform for teams. Its official MCP server makes it the best-integrated cloud platform for AI agent workflows.

Key features: Block editor with 50+ block types; relational databases with views (table, board, calendar, timeline, gallery); real-time collaboration; comprehensive API; official MCP server; built-in AI features.

Pricing: Free tier. Plus $10/mo. Business $18/mo.

Author: Tana Inc.  |  Licence: Proprietary

Outliner meets database for networked thought. Tana’s “supertags” turn any node into a structured object with fields, making it uniquely powerful for combining free-form notes with structured data.

Key differentiator: Supertags (templates that turn notes into structured objects), AI agents that participate in meetings, transcription in 60 languages. In 2026, Tana introduced an agentic system where AI agents autonomously maintain and cross-reference knowledge.

Pricing: Free tier. Pro $9/mo ($84/year) for 5,000 AI credits.

Author: Capacities GmbH  |  Licence: Proprietary

“A studio for your mind” — object-based PKM that feels like a personal database. Instead of pages, you work with typed objects: Books, People, Topics, Projects, and custom types.

Key differentiator: Object-type system provides structure without rigidity. Each object type has its own fields, views, and templates. Calendar integration and robust third-party connections. Particularly strong for managing references and research.

Pricing: Free tier. Pro $10/mo.

Author: Heptabase Inc.  |  Licence: Proprietary

Visual note-taking bridging notes and whiteboards. The canvas view lets you spatially arrange and connect notes, making it ideal for visual thinkers and for mapping complex topics.

Key differentiator: Spatial thinking — drag ideas together on a canvas, draw connections, and see relationships visually. The strongest option for PKM newcomers who find Obsidian’s text-first approach overwhelming.

Author: TOEVERYTHING  |  GitHub stars: 67,100  |  Licence: MIT

Next-generation knowledge base combining docs, whiteboards, and databases — positioned as the open-source Notion + Miro alternative. The “edgeless” canvas mode merges documents with infinite whiteboards.

Key features: Hyper-merged docs and canvas; rich text, sticky notes, embedded web pages, multi-view databases, linked pages, shapes, and slides on the edgeless canvas; local-first (OctoBase engine with CRDT); real-time sync; self-hostable.

Key differentiator: The edgeless canvas mode — documents and whiteboards are not separate features but one unified surface. 67k GitHub stars make it the second most-starred open-source workspace tool.

Author: AppFlowy Inc.  |  GitHub stars: 68,840  |  Licence: AGPL-3.0

The most-starred open-source Notion alternative. Built with Rust (backend) and Flutter (frontend) for exceptional native performance across platforms.

Key differentiator: Native performance — the Rust + Flutter stack makes it significantly faster than Electron-based competitors. The largest open-source workspace community. AI features are pluggable (bring your own LLM).

Author: Outline (General Outline Inc.)  |  GitHub stars: ~29,000  |  Licence: BSL (free for self-hosting)

Fast, beautiful team knowledge base and wiki. The most polished open-source option for teams migrating from Confluence or Notion wikis.

Key features: Notion-like block editor; real-time collaboration; SSO (OIDC, Slack, Google); S3 storage; 2FA; granular access controls; API; webhooks; Slack integration.

Key differentiator: Team-first — while most PKM tools are personal, Outline is designed for team knowledge with permissions, collections, and access controls. Self-hostable (Node.js, PostgreSQL, Redis, S3).

Author: Any Association  |  GitHub stars: 7,368  |  Licence: open source

Object-based PKM with end-to-end encryption and peer-to-peer sync. Every piece of content is a typed “object” with relations, enabling a graph-like content model without cloud dependencies.

Key differentiator: E2E encryption by default with P2P sync (no central server). Object-based model (vs. block-based) allows for richer semantic relationships. Better than Notion for archiving and complex knowledge management where privacy matters.

Author: requarks  |  Licence: AGPL-3.0

Modern, feature-rich self-hosted wiki with the most extensive integration options of any open-source wiki. Real-time collaborative editing (Google Docs-like) sets it apart.

Key features: Multiple editors (Markdown, WYSIWYG, raw HTML); Git-backed storage (use Git as your wiki database); extensive authentication options (LDAP, OAuth, SAML); real-time collaboration; full-text search with Elasticsearch/PostgreSQL.

Author: Dan Brown  |  GitHub stars: 16,000+  |  Licence: MIT

The most popular PHP wiki on GitHub. Organised around a familiar metaphor: Shelves → Books → Chapters → Pages. Low barrier for non-technical users.

Key differentiator: The structured hierarchy (Shelves/Books/Chapters/Pages) provides natural organisation without requiring linking or tagging discipline. WYSIWYG editor makes it accessible to non-technical team members. 5 feature updates + 18 patches in 2025.

Author: Andreas Gohr  |  Licence: GPL-2.0

Lightweight wiki using plain text files with zero database requirement. The simplest possible wiki to deploy, back up, and migrate.

Key differentiator: No database — all content stored as plain text files. Backup means copying a folder. Migration means moving a folder. Plugin ecosystem for extensibility. Battle-tested since 2004.

Author: Meta  |  GitHub stars: 64,400  |  Licence: MIT

React-based static site generator designed specifically for documentation. The most feature-rich SSG for knowledge bases, with native versioning, search, and i18n.

Key features: Built-in Algolia DocSearch integration; native document versioning; internationalisation; React component extensibility; structured sidebar navigation; MDX support (Markdown + JSX).

Key differentiator: The only SSG with first-class document versioning — essential for API docs and knowledge bases that need to maintain multiple versions. React ecosystem means unlimited customisation.

Author: Tom Christie  |  Licence: BSD-2-Clause

Python-based static site generator designed for project documentation. Powers 90,000+ projects. Simple YAML configuration and Markdown files — the fastest path from Markdown to a documentation site.

Key features: Simple YAML configuration; Markdown source files; fast builds; extensive theme ecosystem (Material for MkDocs is the most popular).

Caveat (April 2026): No significant development in 18 months; last release 1.6.1 (August 2024). Material for MkDocs entered maintenance mode November 2025. Successor project “Zensical” announced by the Material author.

Author: GitBook Inc.  |  Licence: Proprietary (free for OSS/non-profit)

Hosted SaaS platform — the easiest path from Markdown to published docs. No build step, no deployment pipeline, no server management.

Key differentiator: Zero ops — push Markdown to a Git repo, GitBook handles the rest. WYSIWYG editing for non-technical contributors. Git sync ensures developers can work in their preferred editor. AI-powered search built in.

Origin: German sociologist Niklas Luhmann (1927–1998)

Luhmann used a physical slip box of 90,000 notes to write 70 books and ~400 scholarly articles. The method treats knowledge as a network of atomic, interconnected ideas rather than a hierarchy of documents.

Core principles:

• Atomic notes — one idea per note, written in your own words
• Unique IDs — every note has a permanent, unique identifier
• Bidirectional linking — notes link to each other, forming a web of ideas
• Emergent structure — organisation arises from connections, not imposed hierarchy
• Bottom-up — structure grows organically from individual notes upward

Tools implementing it: Obsidian, Logseq, Zettlr, Foam, Dendron, The Archive, zkn3

Origin: Tiago Forte (“Building a Second Brain”, 2022)

Organises all digital information by actionability into four categories. Designed for people who need both creative thinking and project execution.

Core categories:

• Projects — active endeavours with a deadline (ship feature X, write article Y)
• Areas — ongoing responsibilities with no end date (health, finances, team management)
• Resources — reference material for future use (design patterns, book notes, research)
• Archives — completed projects and inactive items

Workflow (CODE): Capture → Organise → Distill → Express

Tools implementing it: Notion (via templates), Obsidian (via templates and plugins), Capacities, Tana, any flexible note app

Combined approach: Many practitioners use PARA as the “hub of action” and Zettelkasten as the “spoke of insight” — PARA for organising work, Zettelkasten for developing ideas.