How do you roll out Claude Code to an engineering team?
Short answer
Run a four-week pilot with a few engineers on Team standard seats ($25 a month, $20 billed annually). Write down repo and permission rules, give every agent its own git worktree and tmux session, and measure weekly use and merged work. Then move heavy users to premium seats ($125, or $100 annually). One person per seat.
On this page
Choose the plan and seats
Claude Code comes with Claude's paid plans. For a company, the choice is usually between the Team plan, Enterprise, and paying per token through the API.
| Plan | Price per person | Usage | Fits |
|---|---|---|---|
| Team, standard seat | $25 a month, or $20 billed annually | “more usage than Pro” | Most pilot users |
| Team, premium seat | $125 a month, or $100 billed annually | “5x more usage than standard seats” | People who run agents most of the day |
| Enterprise | $20 per seat a month billed annually, 20-seat minimum, plus usage at API rates | Scales with model and task | Larger orgs that need enterprise controls |
| Pro, Max, Ultra | $20, $100, $200 a month | Individual plans | Solo developers and contractors paying for themselves |
On Team and Enterprise, each member's Claude Code use draws from a per-seat allowance that resets on a rolling five-hour window and a weekly window, shared with Claude chat.3 Start everyone on standard. You will know within two weeks who keeps hitting the limit.
The rollout, step by step
Anthropic's own advice on cost is to “start with a small pilot group and use the tracking tools below to establish a baseline before wider rollout.”3 The steps below follow that.
-
Pick the pilot and an owner
Choose 3 to 8 engineers across at least two teams, including one skeptic. Name one owner who collects the numbers each week and fixes setup problems. Pick real tickets, not demos.
-
Buy seats and decide on overflow
Buy standard seats for the pilot. Decide up front whether people may continue past the allowance with usage credits. On Team and Enterprise, admins can turn usage credits on and set spend limits for the organization, a group or an individual member.3 See How much should you budget for AI tokens per developer?
-
Write the policies down
Keep it to one page: which repos agents may work in, which commands are pre-approved, where secrets live and that agents never see them, who reviews agent pull requests, and what may not be pasted into any AI tool. Put project conventions in a
CLAUDE.mdat each repo root. Organization-wide settings can be delivered as managed settings (server-managed, MDM or amanaged-settings.jsonfile).3 -
Agree a worktree and tmux convention
One task, one branch, one folder, one session. Agents in separate worktrees don't overwrite each other, and a named tmux session survives a closed laptop lid or a dropped SSH connection.
# name = repo-task, same for branch folder and session git worktree add ../api-billing-fix -b billing-fix tmux new -s api-billing-fix -c ../api-billing-fix claudeWhen the pull request merges, remove the worktree with
git worktree remove ../api-billing-fixand end the session. More in How do you run Claude Code in tmux? -
Measure for four weeks, then decide
Use the table below. At the end, upgrade seats that hit the limit most weeks, find out why idle seats are idle, and set the budget for the wider rollout from what you saw.
What to measure in weeks 1 to 4
| Week | Focus | Measure |
|---|---|---|
| 1 | Setup. Everyone signs in on their own seat and finishes one real task. | Seats active, setup blockers, time to first merged change |
| 2 | Daily use on normal tickets | Sessions per person, 5-hour and weekly limit hits |
| 3 | Two or more agents at once, on worktrees | Share of weekly limit used at reset, time agents wait on a person |
| 4 | Review | Merged pull requests with Claude Code, review load, seat changes, budget |
How to collect each number is in How do you measure Claude Code adoption across a team? Expect the trust question to come up: in Stack Overflow's 2025 survey, more developers distrusted AI accuracy (46%) than trusted it (33%).5 Code review rules matter more than a mandate.
Seats, accounts and the terms
Every person gets their own seat and signs in with their own login. No shared logins, and no one person's subscription used by the team. Anthropic's Consumer Terms, which cover individual plans, say:
Anthropic says
“You may not share your Account login information, Anthropic API key, or Account credentials with anyone else. You also may not make your Account available to anyone else.”
Consumer Terms of Service, anthropic.com · Effective Oct 8, 2025 · Checked .
For Team and Enterprise, read the commercial terms your company signs. Neither this page nor Orpheus is legal advice. More on accounts in the terms guide.
Common questions
Standard or premium seats for the pilot?
Start with standard seats at $25 a month ($20 billed annually). After two weeks, move the people who hit the limit most weeks to premium seats at $125 ($100 annually), which carry 5 times the usage of a standard seat.
How many seats can a Team plan have?
The pricing page lists Team for 2 to 150 seats. Enterprise lists a 20-seat minimum.
Can two engineers share one Claude account?
No. Anthropic's Consumer Terms say you may not share account credentials or make your account available to anyone else. Give each person their own seat.
How long should a Claude Code pilot run?
Four weeks is enough to see habits form and to average four weekly resets. One week mostly measures setup.
Sources
- Plans and pricingClaude, undatedChecked
- Anthropic renames Claude Max plans and drops Ultra's five-hour limitThe Next Web, Nov 1, 2026Checked
- Manage costs effectivelyClaude Code docs, undatedChecked
- Track team usage with analyticsClaude Code docs, undatedChecked
- 2025 Developer Survey: AIStack Overflow, 2025Checked