OzBrain for Claude
Give Claude a brain.
Updated
A connector is how Claude talks to your OzBrain. Add OzBrain through the connector menu with the MCP URL. Prefer a computer or Claude desktop for the one-time add. After it is added, Claude can use your brain in its apps, including on your phone. Claude then reads your routing index, pulls the articles a task needs, and stages new knowledge so ChatGPT, Cursor, and Claude Code share the same facts. No API keys and no config files.
From then on Claude stops asking who your clients are, how you write, and what you decided last week. It reads the answer instead. Grant the brain to a collaborator and their agents read it too.
Add the MCP URL, approve once, keep working.
- 01
Open the connector menu
In Claude on the web or desktop: Settings, then Connectors, then Add custom connector. The Claude mobile app cannot add a custom connector. - 02
Add your brain
Paste the MCP URL. Claude discovers the rest on its own.https://ozbrain.com/api/mcp - 03
Sign in and approve
A window opens. You sign in to OzBrain with an email code and approve the connection. Later sessions reuse that approval while Claude keeps a healthy refresh token.
What Claude can do once connected
The connector exposes your brain as tools. Claude decides when to use them, and every write is attributed to Claude in the article history.
Read the routing index
A compact listing of every article and what it covers, so Claude finds knowledge by structure instead of guessing at search terms.
Pull the articles a task needs
Working on a client proposal, Claude reads that client’s article, your positioning, and your voice rules. Nothing else leaks in.
Stage new knowledge
When something worth keeping comes up, Claude writes it to staging, routed to the article it belongs to, never straight into canon.
Surface conflicts before they land
If a write contradicts what the brain holds, the write pauses and Claude sees the conflict. It can record an exception; a different agent or you must approve a supersede.
Also in Claude Code
The same brain works in Claude Code with one command. Your coding agent then knows the same things your chat agent does, which is the point.
claude mcp add --transport http ozbrain https://ozbrain.com/api/mcpKeep Claude. Connect the brain under it.
Projects and chats stay where they are. OzBrain is the structured knowledge underneath, shared with every other agent and the people you grant access to.
Create your brain, follow the guided walkthrough at /start, paste the MCP URL into Connectors, approve once, then use the Claude connector docs for example prompts that exercise different tools.
Questions people ask before connecting
Where are the setup steps and example prompts?
The guided walkthrough at /start covers the click path. The Claude connector docs cover setup, OAuth approval, and example prompts that exercise list, search, read, task, and stage tools.
Does this replace Claude Projects?
No, they stack. A project scopes one workstream; OzBrain holds the knowledge underneath all of them. Read the full OzBrain vs Claude Projects comparison.
Which Claude plans support connectors?
Custom connectors can be added on Claude paid plans from Claude on the web or Claude desktop. The Claude mobile app cannot add a custom connector; after you add it once, Claude can use it on your phone. Claude Code supports the same connection on every plan that includes it.
Can Claude see everything in my brain?
Claude reads through the same scoped tools as every agent, and every read and write is attributed. You can review each article version and who wrote it on the web.
How long does the Claude connection last?
With a healthy refresh token, Claude's connector holds across a normal working day. Day-long use comes from successful refresh, not from a single access token lasting all day. If Claude asks you to reconnect, open Connectors again or revoke and re-approve from your OzBrain account's Connections page. Use https://ozbrain.com/api/mcp