Introducing OpenKnowledge, the best markdown IDE for humans and agents. Open source. Local and private. LLM-wiki ready. Use with Claude, Codex, and your favorite agent today.

1. Try it as a mac Mac app and web ui / cli for other platforms. Try:

2. Highlight feature: a beautiful WYSIWYG editor. Our philosophy was that editing markdown should feel as good as Notion or Google docs for humans.

3. And for agents, 1st class integrations. Built-in side-by-side editing with Claude, Codex, Cursor, Pi, and OpenCode. (We've found it work amazing for writing specs.)

4. For teams, we included sharing natively Backed by git and GitHub -- enables auto-sync so sharing feels closer to Google docs.

5. And made it all open source. We believe a knowledge engine and 2nd brain platform should be open and extensible.

@openknowledge I like this idea, looking forward to trying it. Here is a visualization of the repo, looks like you did a lot of the heavy lifting

@openknowledge let's see if we can integrate with you guys to speed up local rag

@openknowledge There’s embeddings and agentic search in the app, so will def take a look at this as well.

ran a benchmark pointing your semantic search at a local LodeDB endpoint instead of OpenAI. Measuring with your eval harness and wiki it passes the quality gate, query embeds drop 173ms to 4.5ms (as it's local embedding, bonus free and no content egress). Bulk indexing is ~1.8x slower on an 8-vCPU server container, once, in the background. I used the symmetric minilm preset with 384-dim vectors which also cuts the in-process scan and vector cache 4x. This swap alone takes end-to-end search from ~240ms to ~22ms (11x) with zero code changes. I noticed the vector store is not pluggable; benchmarking that separately it tends to hold fine to ~50k chunks; past that the scan hits seconds per query and half a GB of Electron heap (!); LodeDB scans the same shape in ~5ms. That's a separate convo to use a local vector store instead of the current in-process JS cosine

@openknowledge Nice UI, I've been working on a tool to support LLM-wikis too, but it's a CLI and it is mostly focused on frontmatter validation for filtered hybrid search over markdown folders

@openknowledge Nice, yeah, this is something we’re thinking a lot about (linters, rules engines, etc.). Take a look at our MCP design, may be some useful stuff there re:search

@openknowledge What's the collaboration engine based on?

@openknowledge pretty

@openknowledge Way to go. Congrats on the launch!!

@openknowledge great, we need this.

@openknowledge 오 이거 좋나

@openknowledge Local private markdown IDE with agent integrations is great for data sovereignty and security.

@openknowledge I like it.

@openknowledge the "markdown should feel as good as Notion" bit is exactly right. that's the bar most dev tooling doesn't get.

@openknowledge Agree! Was our top requirement.

@openknowledge What’s the difference compared to obsidian?

@openknowledge Obsidian is great, but few things: - WYSIWYG editor (Notion // Google docs like) - Triggering agents (via TUIs) directly from the app - Native MCPs, semantic search, an skills tailored for e.g. LLM wikis Also open source so open to suggestions on how we can go deeper.

@McChickenNoSau1 @openknowledge i use obsidian as well , going to try open knowlegde

@nickgomez @openknowledge Let us know how different it is in the experience!

@openknowledge This is interesting! I was lately exploring the Open Knowledge Format is this related to OKF as well?

@openknowledge We have a starter pack for it! Think of us as a general markdown IDE for any markdown KB. OKF is more of a format for markdown KBs. We don’t restrict to only it, but give tools to help me conformant with it.

@openknowledge That's pretty cool. let me give this a shot!

@openknowledge Let me know if you had any thoughts/suggestions.

@openknowledge 本地私有这点挺关键，Markdown 工具就怕越用越重

@openknowledge Interesting. Question - why should I use this instead of Obsidian?

@openknowledge Your brand is so cute :3 please do release for Android as well.

@openknowledge Interesting, Nick. I've been using Obsidian to work with agents and build my own LLM Wiki. Curious to know how this is different?

@openknowledge Obsidian is great. Here's the tldr of where we focused: - Full WYSIWYG editor (think Notion) - Agent integrations directly within the app - Out of box MCP, skills, and semantic search for LLM wiki support What's most important for you?

@openknowledge Agent integrations directly in the app is interesting. I'll check it out.

@openknowledge The future of AI is built on better knowledge systems. Local, private, and agent-ready is the right direction.

@openknowledge agent loops got 3x faster once we switched to markdown

@openknowledge markdown is the way

@openknowledge I like the lineages viewer for your knowledge graph. It’s so nice to see!

@openknowledge I've been using Obsidian for the past 4-5 years. I'll give @OpenKnowledge a try. Thank you!

@openknowledge Let us know if you have any requests !

@openknowledge please support antigravity and hermes agent

@openknowledge this is exactly the kind of markdown layer i was looking for awesome 👏

@openknowledge honest question: why i should use this instead of obsidian?

Obsidian is great. Here's the tldr of where we focused: - Full WYSIWYG editor (think Notion) - Agent integrations directly within the app - Out of box MCP, skills, and semantic search for LLM wiki support We'll share mor on this soon, but tldr you can use both concurrently as a way to try!

@openknowledge im testing it now and it looks pretty cool!

@openknowledge ?

@openknowledge This is why markdown keeps winning for agent memory. Local files make context inspectable, diffable, and revocable. The missing piece in most setups is expiry: what should the next agent trust, and what should it ignore?

@openknowledge Ooooo 👀

@openknowledge Congratulations 🙌 would love to check this out sooner

@openknowledge Website design is on point

@openknowledge We think good design best indicates good product thinking, so we take it seriously end to end - landing page to the app itself.

@ycombinator @openknowledge Looks interesting, does this work this work like a mind map that we create? Because the ones like graphify and obsidian create patterns. I would prefer creating on one

@ycombinator @openknowledge Yes there’s a graph view! You can use wiki links etc. to link docs together, and the built in MCP and skills help agents automatically do it themselves too

@ycombinator @openknowledge That's really excellent. This will be very useful to me as a lawyer because I now have a more effective method for data retrieval.

@openknowledge Looks great!

@openknowledge For humans and agents” this feels like one of those ideas that seems obvious in hindsight. Nice launch.

@openknowledge setting this up for my local notes tonight

@openknowledge picked this as my "one tool worth stealing" on thursday. local plus llm-wiki ready is exactly the right combo