How to give Claude Code your company's docs

Claude Code is very good at the code in front of it. Ask it why a service was built the way it was, which plan a customer is on, or how your team ships a release, and it has to guess. The answer is in a Google Doc, a Notion page, a Slack thread or someone's head, and Claude Code sees none of them.

This guide covers the ways to close that gap, what each one costs you, and how to set up the one that scales with a team.

What Claude Code can see

In a session, Claude Code works from:

• the files in your repository, and anything it reads or searches there
• CLAUDE.md files, which it loads when a session starts
• what you paste or attach in the conversation
• tools you connect over MCP (the Model Context Protocol)

Everything your company knows that isn't in one of those four places doesn't exist for it. So the question is which of the four should carry your company's docs.

Option 1: put it in CLAUDE.md

CLAUDE.md is the obvious first move: write down what Claude Code should know, and it reads it every session.

It works well for the rules of one repository: how to build and test, the conventions you follow, the things it should never do. It works badly for company knowledge:

• It is read every session, whole. A long CLAUDE.md takes room in every conversation, whether the task needs it or not.
• It belongs to one repository. Your pricing, your customers and your decisions aren't about one repo, so they end up copied into several, and the copies drift.
• Nobody updates it when the source changes. The real copy of a decision lives in a doc or a thread. The copy in CLAUDE.md is correct the day it is written.

Keep CLAUDE.md for the repository, and keep it short. We wrote a separate guide on what goes in CLAUDE.md and what goes in a company brain.

Option 2: paste the docs into the conversation

For a one-off task, pasting a doc works. As a habit it doesn't: every person pastes their own copy, some of them old, the agent learns nothing between sessions, and nobody can see which version an answer came from.

Option 3: connect your company's knowledge over MCP

MCP lets Claude Code call tools while it works. Connect a source of company knowledge as an MCP server, and Claude Code asks it when a task needs it: it searches, reads only what is relevant, and gets back the passages with where they came from.

This is what a company brain is for. Noetix reads your company's website, Google Drive, Notion and Slack, keeps them current, and serves them over MCP to every agent your team uses. Each person's agent signs in as that person, so everyone works from the same knowledge.

Set it up

claude mcp add --transport http noetix https://mcp.noetix.dev/mcp

Tell Claude Code when to use it

Claude Code decides when to call a tool. A few lines in CLAUDE.md make the habit reliable, for example:

## Company knowledge

Our company's knowledge is in Noetix, connected as the `noetix` MCP server.

- Before guessing why something is the way it is, `recall` it.
- For a task with a house procedure (a release, a migration), use `search_skills`,
  then `use_skill`, and follow it.
- If the brain has nothing on a question, `flag_gap` once with the question.

These are the same lines we use in our own repository.

Write procedures down once

A lot of what you want an agent to know is how your team does things: how to ship, how to review, how to record a decision. Write each one as a skill in Noetix. Every connected agent finds the skill that matches its task and follows it, and an admin publishes a skill before any agent sees it. See Skills.

Before you connect a source

Agents can also write back: they can edit a page or start one. Every version of a page is kept and can be restored, and removing anything waits for an admin's approval. See Connect your agents for every tool an agent gets.