Your Brain, in the AI tools you already use.
Webair AI runs your Brain as an MCP server. Connect a client once and it can ask your Brain for context instead of asking you.
Included on every paid plan.
What a connected client can do
Read.Cite.Write back.
Three capabilities, scoped to the request in front of it.
Read
Search your Brain
The client's agent searches your memories, sources and entity pages before it asks you anything.
Cite
Ask and get a cited answer
Instead of raw chunks, it can ask a question and get a synthesized answer with a citation back to where each fact came from.
Write
Save new facts back
What it learns while you work can be written back, so the next tool you open starts from it.
Setup
Approve once.No tokens to paste.
Connecting uses OAuth: you point the client at your Brain, a browser window opens, and you approve it yourself.
Claude Code
One command. Your browser opens and you click Approve.
Claude Desktop
A generated entry in claude_desktop_config.json.
Cursor
A generated entry in ~/.cursor/mcp.json.
Any client that speaks MCP
Point it at the same endpoint. Webair AI generates configs for the three above; the protocol handles the rest.
Headless setup? wbrain_ access tokens still work.
# 1. Point your tool at your Brain
claude mcp add webair-brain \
https://app.webair.ai/api/brain/mcp
# 2. Your browser opens — click Approve.
# No tokens to paste. That's the whole setup.
# 3. Then, inside Claude or Cursor:
> Where did I land with Dana?
⚡ brain_ask("Dana Whitfield")
They're expecting your feedback on the scope document — Section 3, the Cutover Timeline. [Source: scope doc, 2026-03-11]
Cursor and Claude Desktop connect with a config file — both are generated for you in the app.
Permissions
You approve it.You revoke it.
A connected client is a guest with a question, not a copy of your Brain.
Scoped to the request
A client receives context relevant to what it asked — not unrestricted access to everything your Brain knows.
Approved by you
The consent screen is yours. Nothing connects without you clicking Approve in your own browser.
Revoked in one click
Disconnect a client and it stops being able to reach your Brain. Your Brain stays where it is.
Coming from another AI?
Importing memory and connecting over MCP are two different things. You can import what you have already taught ChatGPT, Claude, Gemini or Microsoft Copilot when you sign up — your Brain starts with your history instead of a blank page.
Questions
MCP, answered plainly.
What is MCP?
MCP — the Model Context Protocol — is an open standard that lets AI applications connect to outside tools and data. Webair AI exposes your Brain as an MCP server, so any application that speaks the protocol can request context from it.
Which apps can connect?
Webair AI generates a ready-made configuration for Claude Code, Claude Desktop and Cursor. Any other client that implements MCP can connect to the same endpoint — those three are simply the ones with a generated config and a tested path.
Does connecting a tool give it my whole Brain?
No. A connected client asks your Brain for context relevant to the task and receives that. It does not get a copy of your Brain, and it does not get access to your connected apps or your identity. You approve the connection yourself and can revoke it at any time.
Where does my data live once a tool is connected?
In Webair AI. Connected clients query your Brain over the connection you approved — the Brain itself does not move.
One Brain. Every AI.
Build it in Webair AI, then let the tools you already use read from it.