Claude Code

The terminal agent. One command adds the Brain, one slash command signs you in, and a CLAUDE.md file carries the instructions.

Claude Code is Anthropic's terminal agent. With the brain connected, it reads your company's conventions, decisions and system cards before it touches a repo, and it can file what it learns back into the brain. One command adds the server, one slash command signs you in.

At a glance

Works on	macOS, Linux, Windows (terminal), plus the VS Code and JetBrains extensions
Plans	Any Claude plan with Claude Code access
Sign-in	OAuth from the terminal, opens your browser once
Time	Two minutes
Instructions live in	A `CLAUDE.md` file, not a settings screen

Step 1: add the brain

Run this once, replacing the URL with your MCP URL from the dashboard (Your brain, then Connect AI app):

bash

claude mcp add --transport http lan-brain https://YOUR-BRAIN.brain.lanbrain.ai/mcp --scope user

--scope user makes the brain available in every project on this machine. Drop it if you only want it in the current directory. Use --scope project to share the config with your team through a checked-in .mcp.json:

json

{
  "mcpServers": {
    "lan-brain": {
      "type": "http",
      "url": "https://YOUR-BRAIN.brain.lanbrain.ai/mcp"
    }
  }
}

Start claude and type /mcp.
Select lan-brain. It shows Needs authentication. Press Enter and choose Authenticate.
Your browser opens a LAN Brain sign-in page. Use your usual account and approve.
Back in the terminal the server shows as connected and lists its tools.

The same /mcp screen is where you reconnect if a session ever shows the server as disconnected.

Schematic of the terminal: claude mcp add, then /mcp showing Needs authentication and, after sign-in, connected.

Step 3: put the instructions in CLAUDE.md

Claude Code reads a CLAUDE.md file at the start of every session. Paste the instructions block from the dashboard's Connect AI app tab into one of these:

File	Applies to
`~/.claude/CLAUDE.md`	Every project on this machine (recommended for a personal brain)
`./CLAUDE.md` in a repo	That repo, shared with the team through git (recommended for a company brain)
`./CLAUDE.local.md`	That repo, only you, ignored by git

A minimal company CLAUDE.md looks like this:

markdown

# Working with our company brain

My LAN Brain connector is the source of truth about Acme: its business, people, and decisions.
Before answering anything that depends on my context (my projects, my company,
my decisions, the people I work with), call search_brain first. If the brain has
nothing, say so instead of guessing. When the brain conflicts with what you think
you know, the brain wins; cite the page path.
Do NOT use the connector for general knowledge: definitions, public facts, coding
help, anything you could answer without knowing me.
Answer bottom-line-up-front: the answer first, then only the context that changes
my next step.
When I correct you on something durable, fold it into the brain (apply_learning)
and tell me in one line what you updated. A single ambiguous comment is not a
durable correction.

Use the block the dashboard generates rather than retyping it: it already carries the routing lines if you have several brains.

Try it

"Before we start, read the brain's software conventions for this repo and list what applies."
"What did the team decide about authentication for the mobile app? Link the decision page."
"We just agreed the API version policy in this chat. File it in the brain as a decision under the engineering department."

If something is off