brain

A wiki your agents read before they work and write to when they are done

The one place every chat remembers from.

Pages, files, drafts, versions and comments, in projects with their own people. Claude, Codex and the chat on your phone reach the same artifacts through MCP. What an agent should read first and write back is written on the project's front page, not hard-coded in the app.

Google is the only sign-in, and it is by invitation: someone in an organization invites you first. The link you were sent signs you in.

personal-brand / context.md · published r14 · draft 1 by Claude Code

Personal brand: where things stand

The recurring budget is two hours per week, pessimistic…

Claude Code · Is two hours still the budget?

Přemysl · Keep two, revisit in November.

Artifacts with history

Markdown and HTML rendered in a reading style; every save is a revision with author and summary; nothing is deleted, only superseded. Diff any two.

Drafts and comments

The tools tell an agent to propose a draft and to ask before it publishes; nothing is enforced, so a project writes its own rule on its front page. Threads anchor to a paragraph and stay with it across revisions.

Files, big ones too

Renders, PDFs, 3D models, recordings: uploaded in parts, kept in every version, viewed in place. One bucket, nothing else to run.

Built to be read by agents

Connect Claude Code, Codex, the Claude app or any MCP client with one OAuth sign-in. Every client acts as you, with your access, and is listed under your name. The tools are the same for every project; the conventions are the project's own artifacts.

  1. 1Connect a clientAdd https://brain.flooper.me/mcp and sign in with Google once.
  2. 2Write the project's front pageWhat to read first, what to write back, what never to do. Agents read it before anything else.
  3. 3TalkThe chat already knows where things stand. Drafts and threads come back to you here.
# tools, the same in every project
docs(project) · list_projects() · list_artifacts(project, folder?)
read_artifact(project, path, rev?) · diff(project, path, from, to) · search(query, project?)
save_artifact(project, path, content, summary, publish?) · publish(project, path, revisionId) · append(project, path, text)
list_threads(project, path) · open_thread(project, path, body, anchor?) · comment · resolve_thread
start_upload · upload_part · finish_upload · list_files · get_file_url

# what an agent does with them is the project's front page:
"Before working, read context.md and the last week of journal.md.
 After working, append one paragraph to journal.md. Publishing is mine."

Organizations, projects, people

An organization holds projects and people. A project is a wiki with its own access: everyone in the organization, or selected people only. Any single page or file can be shared by a secret link; nothing else leaves. Guests see only the projects they are invited to. Alone, you are an organization of one and nothing of this gets in the way.

Owner
Everything, including people, roles and deleting the organization.
Admin
People, invitations, roles, project access and settings.
Member
Sees every project set to "everyone in the organization"; edits where granted.
Guest
Only the projects they are invited to, with the role given there.

Questions

Is it a note-taking app?
No. It is the memory a person and their agents share: pages with history, files with every version, comments where the text is, and a front page per project that says how work runs there.
What can an agent do?
Everything the person who connected it can: read, search, draft, publish, comment, upload. The tools say what each one does and that the agent should ask before publishing; a project writes its own rule on its front page.
Who can see what?
An organization's owners, admins and members see the projects open to the organization; guests see only the projects they are invited to; nobody outside sees anything, except single pages shared by a secret link.
Where does it run?
On Cloudflare: one Worker, one database, one bucket. Every file revision is kept; the database is exported nightly and a restore is tried weekly.
Can I take my data out?
Any project exports as plain files and a history of every revision. Pages are markdown; nothing is locked in.
brain · a wiki for people and their agentsSign in