LAN Brain docs
Documentation for connecting your AI to your company brain and getting real work out of it. Start with Connect your AI: one guide per client. Guides cover conventions, best practices, and keeping the brain healthy. The endgame stages cover how to automate, iterate, and maintain the brain over time.
Connect your AI
One guide per client, with the exact steps, the config to copy, and the instructions that make your AI actually use the brain.
- ClaudeClaude.ai on the web, the desktop app and mobile. One connector, added once, follows your account everywhere.WebDesktopMobile
- Claude CodeThe terminal agent. One command adds the brain, one slash command signs you in, and a CLAUDE.md file carries the instructions.TerminalIDE
- ChatGPTCustom connectors in developer mode. Business, Enterprise and Edu get full support; Pro gets read access. Web only.Web
- GeminiThe Gemini app (US personal accounts), the Gemini CLI for developers, and Gemini Enterprise through your Google Cloud admin.WebCLIEnterprise
- CursorOne JSON entry in mcp.json, global or per project, and your coding agent reads the company brain while it works.Editor
- VS Code and CopilotCopilot agent mode reads the brain through .vscode/mcp.json or the MCP: Add Server command. OAuth is built in.Editor
- Windsurf, Zed and othersWindsurf, Zed, Perplexity, Microsoft 365 Copilot, and any client that speaks MCP over HTTP.EditorWebEnterprise
- Custom instructions
Connecting is not adopting. The one block of text that makes your AI reach for the brain, why it is short, and where to paste it in each client.
- Real-world use cases
Twelve things teams actually do with a connected brain, each with the prompt that starts it: onboarding, call prep, weekly reports, decisions, code context.
- Good habits
How people who get the best answers ask, correct, and keep the connection healthy. Also what not to do.
- Troubleshooting
The connector does not appear, the AI answers from memory, sign-in loops, empty results. Each symptom with its fix.
- Security and access
What your AI can and cannot see, how sign-in works, what is logged, and how to cut access in one step.
Guides
- How it works
The technical surface: how your AI reaches your brain over MCP, the session bootstrap and four core files, structured descent instead of vector search, the tool list, the Systems Registry, and skills.
- Conventions
Naming, folder structure, wikilinks, frontmatter, and aliases. The short reference for how things are written inside your brain.
- Best practices
The weekly and monthly rhythms that keep a brain growing. Atomic notes, entity disambiguation, archive versus delete.
- Backup routine
LAN snapshots your brain daily, so backups are not your job. What replaces the routine: reviewing recent changes, keeping integrations connected, keeping the brain curated.
- Health checks
Weekly linter runs, monthly graph walks, quarterly skill audits. The signals that tell you a brain is healthy or quietly rotting.
- Learning paths
Three roles inside one company brain: the curious newcomer, the power user, the operator / admin. Pick the one that fits today and move on when it stops being useful.
- What to avoid
Heavy files, duplicate data, broken wikilinks, never reviewing what changed, pasted secrets. The nine mistakes that hurt brains the most.
Endgame
- Automate
Let the brain do the recurring work. Skills are saved procedures your AI runs on demand, and a standing catalog handles the routines: sweeping sources, weekly reviews, change oversight.
- Iterate
The brain improves through use. Ask real questions, review what the AI writes back, correct what is wrong. Every change is reversible, so you can experiment freely.
- Maintain
Keep the brain current and trustworthy. The AI does the upkeep through the skills catalog; you review, and check that the sources are still connected.