Overview
SiYuan is a privacy-first personal knowledge management system built around block-level references, bidirectional links, and Markdown WYSIWYG editing. It combines notes, databases, flashcards, web clipping, PDF annotations, APIs, a CLI, and Docker deployment in one workspace, while the current source tree also includes agent and MCP subsystems. Users can keep data in their own workspace while choosing native desktop/mobile apps or browser-oriented self-hosting.
Features and best fit
Based on official documentation; not hands-on tested · Content checked:
Treat documents as fine-grained knowledge blocks with references, backlinks, and Markdown WYSIWYG
The official README describes SiYuan as a personal knowledge management system with block-level references and Markdown WYSIWYG. Its editor combines block references, bidirectional links, custom attributes, SQL query embeds, block zoom, table views, flashcards, formulas and diagrams, web clipping, and PDF annotation, allowing information to be composed as reusable knowledge blocks rather than only page-sized notes.
Sources: [1]
Keep the workspace local while exposing the same data through Docker, APIs, and a direct CLI
SiYuan stores data under the workspace data directory, with document data represented as .sy JSON files in notebook folders. The README documents Docker deployment and APIs as well as a CLI that can access workspace data without a running server, covering notebook/document/block CRUD, full-text and semantic search, asset search, SQL queries, Markdown export, history, and sync operations.
Sources: [1]
Extend AI-assisted writing toward an agent runtime with MCP-connected tools
The stable feature list includes AI writing and Q&A through the OpenAI API. The current repository also carries agentic-ai, ai-agent, and mcp topics and contains kernel/agent and kernel/mcp implementations. The agent runtime prepares MCP server connections for agent requests, showing an expansion from AI-assisted editing toward agents that can operate against the knowledge workspace through tools.
A strong fit for block-oriented PKM users who also want self-hosting, automation, and agent access
SiYuan is relevant when locally controlled notes need finer-grained references than page-only links, while databases, flashcards, web clipping, and PDF workflows should remain in one system. It is also useful to developers who want CLI/API access to the workspace or want to experiment with AI agents and MCP operating against a personal knowledge base.
Synchronization is not generic folder sync and depends on SiYuan's data repository and supported providers
The official FAQ explicitly says that synchronizing the workspace through a third-party sync-disk folder is unsupported because it may corrupt data. SiYuan's own synchronization can use official cloud or third-party storage such as S3 and WebDAV, but third-party cloud-storage connectivity is a member privilege. Treat the SiYuan data repository and supported sync providers as the synchronization boundary rather than assuming any folder-sync tool is safe.
If the data-repository key is lost and cannot be recovered from another configured device, the documented recovery flow requires a manual backup and reinitialization of the data repository; existing cloud snapshots can no longer be used. Maintaining a separate backup of workspace/data is therefore prudent.
Sources: [1]
Docker is browser-oriented and requires deliberate authentication, volume persistence, and AGPL review
The official Docker path exposes the b3log/siyuan image on port 6806, mounts a host workspace, and requires an accessAuthCode. The README explicitly warns users to change that code and to verify the mounted volume or risk losing data when a container is removed. Docker hosting does not support desktop/mobile app connections, PDF/HTML/Word export, or Markdown import.
The source is AGPL-3.0 licensed. Redistribution, modification, and network deployment of modified versions should be reviewed against the AGPL obligations.
Official sources
- [1]siyuan-note/siyuan — README(2026-09-15)
- [2]SiYuan kernel — Agent runtime(2026-09-15)
- [3]siyuan-note/siyuan — LICENSE(2026-09-15)
Supplemental curator note
We selected SiYuan because it treats blocks as the core knowledge unit while bringing GUI, CLI, APIs, and agent/MCP access onto the same workspace. Before adoption, verify that generic sync-folder tools are unsupported, S3/WebDAV sync is a member feature, and Docker hosting has functional limitations.
Try it in 3 steps
- 1
Create a persistent workspace
Create a host-side workspace so your data survives container removal.
mkdir -p "$PWD/siyuan-workspace" - 2
Start SiYuan
This follows the official Docker deployment pattern. Replace accessAuthCode with your own strong value before use.
docker run -d --name siyuan -v "$PWD/siyuan-workspace:/siyuan/workspace" -p 6806:6806 b3log/siyuan serve --workspace=/siyuan/workspace --accessAuthCode="replace-with-a-strong-secret" - 3
Open it in the browser
Sign in with the access code, create a notebook/document, and try block references and Markdown WYSIWYG editing.
http://localhost:6806
Growth
Growth trends · Last 30 days
46,372 Stars
Trend data is still being collected.
Development activity
Last 90 days · weekly
- Commits (last 30 days)
- 786
- Open PRs
- 1
Development activity is still being collected.
Built with
Categories and tags
GitHub data
GitHub dataView detailed GitHub data
GitHub Topics
- note-taking
- local-first
- knowledge-base
- markdown
- self-hosted
- siyuan
- ai-agent
- mcp
- agentic-ai
- Stars
- 46,372
- Forks
- 3,003
- Watchers
- 188
- Open issues
- 14
- Primary language
- TypeScript
- License
- AGPL-3.0
- Repository last updated
- Sep 15, 2026
Explore next
- OpenClaw389,730 Stars
A persistent personal AI assistant platform spanning channels, tools, memory, and schedules.
TypeScript - Hermes Agent245,655 Stars
A persistent AI agent with memory, skills, cron jobs, subagents, and multiple terminal backends.
Python - n8n204,350 Stars
A fair-code platform combining a visual canvas and code for service integrations, workflows, and AI agents.
TypeScript - TensorFlow200,091 Stars
An end-to-end machine-learning platform connecting research, training, evaluation, visualization, and deployment.
C++ - Ollama181,014 Stars
A runtime for downloading and running open models on a computer or server through a CLI and local API.
Go
Report incorrect information
Tell us if any listing information is incorrect or outdated.