Nick Gomez

founder @openknowledge @inkeep

prev. @microsoft @mit

[Nick Gomez @nickgomez]

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.

[Nick Gomez @nickgomez]

@jayair @orca_build does all of this pretty well

[Nick Gomez @nickgomez]

@cold_front17 @lody_ai @openknowledge @inkeep also checking in on this from OpenKnowledge side.

[Nick Gomez @nickgomez]

@RLerelou @openknowledge We recently added Excalidraw support, but yes, we have a few things in this space cooking

[Nick Gomez @nickgomez]

@RLerelou @openknowledge It's fully "self-hosted" (on your laptop/computer) // nothing goes through our servers so risk vectors are mitigated from a centralized server standpoint.

As we work on the cloud offering we'll go through pen tests and other typical audits.

[Nick Gomez @nickgomez]

@Andrew652263 @stewnight @openknowledge Haven't personally tried it yet but tldr we expose the mcp // ui over a port, it should be possible to use any type of proxy/networking tool to add whatever authn layer you like.

[Nick Gomez @nickgomez]

@openknowledge Keep up the good work guys! I was setting up an Obsidian llm wiki when i saw your project. With what you launched it’s way better than anything i could find based on Obsidian.

In terms of security though will you get an audit?

[Nick Gomez @nickgomez]

@m31uk3 embedded HTML already there ! (though more coming on this soon).

curious what you have in mind re:interactive walkthroughs?

[Nick Gomez @nickgomez]

LLM wikis now have a standard format: the Open Knowledge Format (OKF).

Today we released our open-source plugin to make it easy to create and maintain wikis in this format. It's part linter, part skill, part MCP tools.

OKF was introduced by Google back in June as a specification that formalizes the LLM wiki concept.

What's elegant is that the rules are progressively opt-in, it's relatively easy to convert any collection of markdown files into a proper OKF-conformant knowledge base.

Underneath, your knowledge base is just markdown and yaml files, so it's not a proprietary format that only certain clients or apps can read from.

The spec is additive, helping agents navigate and edit the wiki more easily, but it stays portable and interoperable.

A lot of the same principles to how and why we built OpenKnowledge, our markdown IDE.

Our new plugin enables lint and deterministic rules for auditing your knowledge base, as well as skills and MCP tools that agents leverage as they write to stay conformant.

The way it works is neat: agents have the skill loaded as they write. Still, if they make mistakes, they get instant feedback from the 𝚠𝚛𝚒𝚝𝚎 MCP tool response as a warning, so it can fix them immediately.

They also have access to 𝚊𝚞𝚍𝚒𝚝 and 𝚕𝚒𝚗𝚝 tools to check and fix issues at any time, which is useful for migrating any existing knowledge base to follow the format.

More details in the blog.

[OpenKnowledge @openknowledge]

Make any knowledge base or LLM wiki follow the Open Knowledge Format with our new OKF plugin.

OKF is an open spec created by Google that makes markdown-based knowledge bases portable and interoperable through a set of shared patterns and rules.

With the new plugin, you can easily convert any existing knowledge base to OKF, and have agents write and maintain your wiki in the format.

The plugin leverages a combination of linting, deterministic rules, and automations for creating special files like 𝚒𝚗𝚍𝚎𝚡.𝚖𝚍.

It also adds a skill and leverages our MCP tools to give agents real-time hints and warnings as they write, so issues get caught immediately — plus 𝚊𝚞𝚍𝚒𝚝 and 𝚕𝚒𝚗𝚝 tools to check the entire knowledge base any time.

[Nick Gomez @nickgomez]

@EadrictheWild @openknowledge Let me know if you have any feedback. dms open.

[Nick Gomez @nickgomez]

@stewnight @openknowledge Got it. Let me know if you have any issues, dms open.

[Nick Gomez @nickgomez]

OpenKnowledge MCP has a virtualized tooling system where we enrich cli tools like `ls` with read/write operations for agents with a list of backward links and other metadata.

We find this helps agents navigate and find the right content without needing vector search. Some folks call this hierarchical search or agentic search. Works for vast majority of KBs.

That said, things can be combined -- e.g. you can add vector embeddings (including with OSS embedding models) and that'll enrich the `search` tool.

[Nick Gomez @nickgomez]

@stewnight @openknowledge Glad to hear !

For auth: check out recent new docs on this around Remote Control, etc.

We recommend adding something like tailscale/ngrok.

https://t.co/ePNX5AsC2A

[Nick Gomez @nickgomez]

@gmccreight Super neat, will check it out. Publishing // pretty rendering wikis is something we’ve given some deep thought to.

[Nick Gomez @nickgomez]

@evalstate Nice, good to see the ecosystem pick up. Particularly useful for research.

[Nick Gomez @nickgomez]

@blumbuilds This is true. We do keep edit history and expose that to agents so that provenence is clear, and OKF .2 introduced sources which in theory helps with citations.

That said, semantic conflict resolution is the most interesting and imo open research problem in the space.

[Nick Gomez @nickgomez]

@shinzui Neat. I was very surprised Google didn't publish some type of linter or tooling around OKF when introduced.

[Nick Gomez @nickgomez]

@___frye @openknowledge ~~ open source wysiwyg editor for markdown with team sharing

https://t.co/ZeXnVGeXA5

[Nick Gomez @nickgomez]

@_ocooper agree. imo virtually any asset should be represented in markdown, html, css and react. i think that's where the world is going.

[Nick Gomez @nickgomez]

We run this same filter every time we make a choice on infra, dependency, or dev tool: is it open source.

It's not just because it helps agents figure out "what's wrong" when something isn't working, but also because it lets agents test things end to end.

Especially when we're making choices about what goes into our product, it's virtually a hard requirement for us that whatever dependency we choose allows us to still fully re-create the full environment our real product runs on on-demand, infinite times.

This is needed so that agents and agent loops can test changes and new functionality end to end to the highest fidelity possible.

Everything needs to be 100% reproducible so that agents can you actually have a validation loop and prove their work.

[Rhys @RhysSullivan]

i love using open source software now because whenever something isn't working as expected apparently my agent is able to just go look at the source code of it and tell me how it works

[Nick Gomez @nickgomez]

Right now we do support a) pi agent in the chat UI b) do have a conversation history as part of the ux (for chats that happen within OpenKnowlefge)

It's not as full chat-session first experience as chat-first ADEs like codex/conductor., but for writing specs, artifacts, or knowledge work it works well.

[Nick Gomez @nickgomez]

@BenLissen @openknowledge Ah, noted, we'll take a look.

[Nick Gomez @nickgomez]

One of the key differences between @openknowledge and Google Docs and Notion is that you can "bring your own agent".

I think all SaaS and desktop apps in the long run will converge towards this model.

The reason: people have a preference on their preferred agent and don't want to be switching between per-vendor agents.

They want agents to have their skills, MCP servers, and file system context (like your code base) without re-hooking it up all of those things to yet another service.

In company environments, the agent available to you is often predetermined by IT. And for personal use, you probably want to use your existing subscriptions.

With our launch today, you can now chat with 35+ agents directly within the OpenKnowledge app. This is thanks to the ACP protocol that standardizes how to interface with agents across 35+ harnesses.

It let us built a native UX experience, with unique features like like Follow Mode and Comments.

And for folks who prefer to work from their existing UI experiences, we kept the built-in terminal "TUI" agents as well and the Claude/Codex desktop app integrations.

OpenKnowledge = edit and view your content from and with any agent.

[OpenKnowledge @openknowledge]

We now support 30+ agent harness integrations within OpenKnowledge.

This is enabled by the Agent Client Protocol (ACP), which let us build our custom UI experience while fully leveraging the agents on your computer.

Also powers our new Follow Mode and Comments features.

Follow mode automatically navigates the IDE to the files being edited by agents, and animates agent changes so it's easy to follow along.

Comments let you annotate and mark up specs, blogs, and other documents and send all the feedback in one go to agents.

Now enabled in the Mac, Linux or Windows apps.

~Day 3/5 of Launch Week~