Agent brain
A shared knowledge base your agents read from and write to — the persistent memory behind every session.
A rich text editor for you and your agents. Private, open source, and free.

v2.0 went public on June 3 — the end of a quiet QA window and the start of launch week.
47 PRs merged across the cycle. Activity stayed close to zero through QA, then spiked on launch day as the announcement went live.
v2.0 went public on June 3 — the end of a quiet QA window and the start of launch week.
47 PRs merged across the cycle. Activity stayed close to zero through QA, then spiked on launch day as the announcement went live.
v2.0 went public on June 3 — the end of a quiet QA window and the start of launch week.
47 PRs merged across the cycle. Activity stayed close to zero through QA, then spiked on launch day as the announcement went live.
A Notion or Google-docs editor experience that's just markdown under the hood.
Pin key context next to the section that needs it.


Highlight a tip, warning, or aside inline.
Collapsible section with a clickable summary.
Switch between content panes inline.
Render flowcharts and diagrams from text.
Drop in images with captions and resize controls.
Embed videos with playback controls inline.
Tables, images, code, diagrams, and more.
Drop-in apps, previews, and visualizations.
Ready-made and customizable templates.

I'll capture this as a decision doc.
Done — edit landed:
Built-in tools for agents to co-author and understand your knowledge base.
Your agents know how to navigate, edit and grow your knowledge base out of the box.
Plug your docs straight into Claude, Cursor and Codex. No copy-paste, no export.
Every change is yours to keep, tweak or undo. Nothing is final until you say so.
A shared knowledge base your agents read from and write to — the persistent memory behind every session.
Specs, RFCs and runbooks living next to the code, edited by humans and coding agents in the same loop.
A living wiki for your team — docs, notes, and decisions in plain markdown, kept current by everyone who works on them.
We’re bringing the whole team into one shared workspaceClaude, where people and agents edit the same files instead of losing context in a thread.
One click to share a doc or a whole space. People and agents collaborate on the same files; everything syncs through git, so history and ownership stay yours.
Local-first, open source, plain files you own. Platform-agnostic by design.
Runs on your machine; private by design.
Auditable, self-hostable, & community-built.
Use with any agent and any tool.