# Run parallel Claude Code agents on one repo

<!--
title: "Run parallel Claude Code agents on one repo without conflicts · Cork"
description: "Cork lets several Claude Code agents work on one repository at once: each starts from a shared map of the code, hears the others' changes on its next step, and has a write held when it would break theirs. One command, free."
canonical: "https://corkcli.com/use-cases/parallel-claude-code-agents"
last_updated: "2026-11-15"
-->

Install Cork once, then open as many Claude Code sessions as you like. Each one starts from a shared map of the code, hears the others' changes on its next step, and lands with one git push.

Install, free and local, no account: `npx corkcli` (or `curl -fsSL https://corkcli.com/install.sh | sh`). MCP only: https://mcp.corkcli.com

## How it works

1. **Install once.** `npx corkcli` sets up the hooks for every repo on the machine. No account, nothing written into your repos.
2. **Start the agents.** Open Claude Code sessions as usual, in one checkout or in worktrees. They need no new instructions.
3. **Let them work.** Briefings, teammates' changes and held writes arrive in each session's context. `git push` rebases, tests and pushes.

## Measured

**47%** fewer tokens with 16 Claude Code agents on one repo, five rounds each way. [Method and raw numbers](https://corkcli.com/benchmarks/study-7-scaling-to-sixteen-agents)

## Questions

### Do I need git worktrees to run Claude Code agents in parallel?

No. Cork works with every agent in one checkout or each in its own worktree: a repo is known by its git remote, so every worktree of it is the same repo to Cork. Worktrees keep the files apart; Cork keeps the agents aware of each other either way.

### Does Cork slow Claude Code down?

The hooks run on your machine and make no model calls. Every hook fails open: an error, a timeout or a missing config lets the action through.

### Is it different from Claude Code's subagents or agent teams?

Subagents and agent teams split one task inside one session. Cork coordinates separate sessions, yours and your teammates', on one repo, whatever started them.

### What does it cost?

Cork on one machine with no account is free. Signed in, 10 agent hours a month are free (30 more the first month); after that it is $20 a month per 100 agent hours, pooled across the team. An agent hour is time the model is thinking and writing; builds, test runs and idle time never count.

## Read more

- [Cork for Claude Code: setup](https://corkcli.com/docs/claude-code)
- [Run multiple Claude Code agents on one repo](https://corkcli.com/blog/run-multiple-claude-code-agents-one-repo)
- [Do you need git worktrees for parallel agents?](https://corkcli.com/guides/do-you-need-git-worktrees-for-ai-coding-agents)
- [Subagents vs parallel sessions](https://corkcli.com/blog/claude-code-subagents-vs-parallel-sessions)

[All use cases](https://corkcli.com/use-cases.md) · [Cork in Markdown](https://corkcli.com/index.md) · [llms.txt](https://corkcli.com/llms.txt)

This page in HTML: https://corkcli.com/use-cases/parallel-claude-code-agents
