[ integrations / claude-code ]

Team conventions for Claude Code

Claude Code reads project skills and MCP servers. Codehabits generates both from how your team reviews and ships — so Claude Code stops inventing patterns your reviewers would reject.

How it works

  1. 01

    Baseline + PR intelligence

    codehabits enable analyzes the repo and optional PR history into structured .codehabits/ JSON.

  2. 02

    Skills Claude Code discovers

    Skills are written under .claude/skills/codehabits-team-intel/ by default, plus shared .agents/skills paths.

  3. 03

    MCP for active checks

    Add the Codehabits MCP server with Claude's MCP CLI so tools are available in-session.

Setup

Step 1

Enable Codehabits

Creates .codehabits/ and generates skills Claude Code can load.

npx @codehabits/cli enable

Step 2

Register the MCP server

Adds Codehabits tools to Claude Code.

claude mcp add codehabits npx -y @codehabits/mcp

Step 3

Restart Claude Code

Confirm MCP tools appear, then ask Claude to respect .codehabits/ conventions.

What you get

  • .claude/skills/codehabits-team-intel/ (default skill path)
  • AGENTS.md Team Intelligence section
  • .codehabits/ structured intelligence files
  • Eight MCP tools including get_team_context and check_code

Docs & guides

FAQ

Where do Claude Code skills live?
Codehabits writes to .claude/skills/codehabits-team-intel/ by default, and also shared .agents/skills paths so other agents see the same intelligence.
Can I use Codehabits without MCP in Claude Code?
Yes. Skills and AGENTS.md provide passive context. MCP is optional for live queries and feedback loops.