Claude Code for Teams: Plans, Shared Setup and Agent Teams
To run Claude Code across a team, put everyone on a Claude Team or Enterprise plan, check a shared CLAUDE.md (or AGENTS.md) and .mcp.json into the repo so every session starts with the same rules and tools, give each agent its own git worktree, and review through pull requests. Claude Code’s agent teams coordinate one person’s sessions; for several people’s agents, add a shared thread.
Key points
| Plans with Claude Code | Pro and Max (individuals); Team (2 to 150 people); Enterprise |
| Team plan seats | Standard $20/seat/month billed annually ($25 monthly); Premium $100 ($125 monthly) |
| Enterprise | $20/seat/month billed annually, 20-seat minimum, usage billed at API rates |
| Shared instructions | CLAUDE.md in the repo, read at the start of every session; AGENTS.md when there’s no CLAUDE.md |
| Shared tools | Project-scoped MCP servers in .mcp.json |
| One person, many agents | Subagents; agent teams (experimental, off by default) |
| Many people | Git worktrees and PRs; a shared thread such as elvex |
1. Pick the plan
Claude Code comes with Claude’s paid plans. The Team plan is for 2 to 150 people: a standard seat is $20 a user a month billed annually ($25 monthly), and a premium seat with higher limits is $100 ($125 monthly). It includes Claude Code and Claude Cowork, with single sign-on. Enterprise is $20 a seat a month billed annually, with a 20-seat minimum and usage billed at API rates, and adds SCIM provisioning, audit logs, a Compliance API and a HIPAA-ready configuration. Developers can also use Claude Code with a Claude Console account, paying API prices.
2. Share the setup through the repo
- CLAUDE.md: a Markdown file at the project root that Claude Code reads at the start of every session. Put your conventions, commands, and do’s and don’ts there, and review changes to it like code.
- AGENTS.md: Claude Code reads it when there’s no CLAUDE.md, and can be set to read both. If part of the team uses Codex, which reads AGENTS.md, keep the shared rules in AGENTS.md and point CLAUDE.md at it.
- MCP servers: add shared servers at project scope (
claude mcp add --scope project …). They’re saved in.mcp.jsonin the repo, and each person approves them before first use. - Subagents: custom subagents are Markdown files in
.claude/agents/, so a team can share a reviewer or test-writer agent through the repo.
The full list of what you can and can’t share is in how to share Claude Code with your team.
Claude Code agent teams: what they are and when to use them
Agent teams are an experimental Claude Code feature, off by default, in which a lead session spawns teammate sessions, assigns them tasks and merges the results. All the sessions belong to one person. Use them to split one person’s large task across several Claude Code sessions; they don’t connect two people’s agents.
To try them, Anthropic documents setting CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1.
3. Subagents vs agent teams
| Subagents | Agent teams | |
|---|---|---|
| What it is | Specialized helpers inside your session, each with its own context window, tools and permissions | A lead session that spawns and supervises teammate sessions and assigns them tasks |
| Status | Generally available | Experimental, off by default |
| Who it serves | One person | One person |
Both run several agents for one person. Neither puts two people’s agents together. See Claude Code subagents.
4. Keep agents from colliding
Give each agent its own git worktree and branch, so two sessions never edit the same checkout, and make the pull request the hand-off point. Claude Code can also run as a cloud session in an isolated VM, keep going after you close your laptop, and watch a pull request to respond to CI failures and review comments.
5. Review and permissions
Claude Code’s permission modes run from read-only to auto, where a classifier reviews actions, and Bash commands can run in an OS-level sandbox (/sandbox). Agree as a team which mode to use where, and keep human review on the pull request.
6. Coordinate several people’s agents
Worktrees keep changes apart, but the agents still don’t know about each other. When Priya’s Claude Code is about to change a function that Diego’s agent depends on, someone has to notice and relay it. Claude Code has no built-in multi-user session (see Claude Code multiplayer).
elvex is a shared thread where each person’s agent reads the same plans, claims and hands off to-dos, flags conflicts before they land, and brings a person in only for decisions. Each person adds it to their own Claude Code:
claude mcp add --transport http --scope user elvex https://api.elvex.ai/mcp
claude mcp login elvex
Teammates on Codex or Cursor join the same thread. Setup for every client is on the elvex MCP page.
Frequently Asked Questions
Does Claude Code have a team plan?
Yes. The Claude Team plan, for 2 to 150 people, includes Claude Code: standard seats are $20 a user a month billed annually ($25 monthly), and premium seats $100 ($125 monthly).
What are Claude Code agent teams?
An experimental feature, off by default, in which a lead Claude Code session spawns teammate sessions, assigns them tasks and merges the results. It coordinates one person’s sessions.
How do I share Claude Code settings with my team?
Commit CLAUDE.md (or AGENTS.md), project-scoped MCP servers in .mcp.json, and custom subagents in .claude/agents/ to the repo.
Can my whole team work in one Claude Code session?
No. Each session belongs to one person. To have several people’s agents work together, connect each one to a shared thread such as elvex.
Sources
Claude Code overview, memory (CLAUDE.md and AGENTS.md), subagents, MCP, security, Claude Code on the web and Claude pricing (Anthropic), checked November 20, 2026.