---
source: CrewKit integrations
url: https://crewkit.io/integrations/index.md
title: "Integrations"
description: "How CrewKit works with Claude Code, Codex, Cursor and any MCP-capable agent."
---


CrewKit connects the AI coding agents your team already runs to one shared
project context: the same agents and rules, the same conventions, the same
memory, and one place where every session is visible.

## CrewKit for Claude Code

The deepest path. crewkit code launches the native claude binary and observes it
through the hook system, so turns, tokens, cost, subagent tasks and post-session
summaries all reach the dashboard.

It is the only integration where the work is measured, not just informed — the
Claude Code experience itself does not change.
[Read the guide](/integrations/claude-code).

## CrewKit for Codex

Standards, not telemetry. Playbook conventions — category, key, text and
anti-patterns, merged across the platform, organization and project tiers —
reach Codex through the get_conventions MCP tool.

A second agent does not mean a second set of rules.
[Read the guide](/integrations/codex).

## CrewKit for Cursor

Memory in the editor. The documents, transcripts, captured messages and analyzed
past sessions that make up project memory stay searchable from the editor
session, next to the code they explain.

The reason a module is written the way it is stays reachable while you are
editing it. [Read the guide](/integrations/cursor).

## CrewKit for any MCP client

The general door. Nine tools over stdio, scoped to one organization and one
project, for any client that speaks the Model Context Protocol — including the
ones not listed here.

Register the server by hand and the same project context answers.
[Read the guide](/integrations/mcp).