One brain every agent reads and writes.

The work that chats drop.

A finished decision, a plan, a piece of research, and how the person wants things done belong in articles, not in a single thread. Host memory keeps scraps. Files multiply across chats, repos, and docs until nobody can tell which copy is current. OzBrain holds one canonical version. A new conversation, or a different agent, reads the same articles. When something is out of date, freshness on the article says so. When the person has been repeating a preference, that preference is an article the next session can read instead of being told again.

What changes when that work is an article.

1. A decision, plan, or document that vanished from a thread is an article the next session can read.
2. A preference, or how the person wants things done, is an article instead of another explanation.
3. When an agent cannot tell whether something is still true, the write goes in an article whose freshness says whether it is current.
4. The same work in a chat, a file, a repo, and a doc becomes one canonical article.
5. At the start of work the agent reads the brain, so the person does not paste a brief, re-explain a project, or move files between tools to keep the agent informed.
6. A new conversation, or another connected agent, reads the same article. The thing that mattered does not have to be retold for it to survive.

The first article.

The first article is who the person is and how they work because later articles are written about that person, and every other connected agent needs the same profile. It is compiled from what a connected agent already knows, then corrected before it is saved. Later articles are the work itself. Each change is staged. Promotion makes it the current version the person approved. The agent that writes it does not keep a private copy that other agents cannot see.

After connecting at https://ozbrain.com/start, a person pastes this into the connected agent:

Compile a profile of me from what you already know in this host: memory and past conversations. Cover my name, location, role, and how I work, plus any other details that help when writing about me or my work. If you do not have those basics, ask me before you finish the draft.

Present the draft as: this is what you believe about me; some of it is probably wrong; correct it before you save. Wait for my corrections before saving.

After I correct it, save the profile as an article in OzBrain. Mark the change as an import (for example import: create profile). Show me the staged write, then make it the current version when I approve.

Connected agents read, write, and coordinate.

Read

Connected agents list the brains a connection can see, open a brain routing index and its instructions, inspect recent activity, list tasks, open a task, read an article with its provenance, search articles, check a pending write, and export the brain as markdown.

Write

Connected agents create a brain, stage an article change, promote a staged write to the current version, record a migration follow-up, verify an article, import a markdown bundle, and set links between articles.

Coordinate

Connected agents create a task, claim the next task, claim a specific task, keep a claim alive, release a claim, force-release a claim, update task status, update a task, approve a task, cancel a task, comment on a task, attach a link to a task, connect a task to an article, and propose inviting someone to a brain.

Plans and limits

Free is $0 forever, up to 50 articles. Pro is $20 per month, up to 300 articles. Max is $99 per month, up to 600 articles. Company is Custom, talk to us. Org-owned brains every seat's agents share. Reads, writes, connections, and brains are not metered.

Connect

People connect at ozbrain.com/start. After a connection exists, the first write is a staged article. Promotion makes it the current version. Claude connector setup is at ozbrain.com/docs/claude-connector.