Claude Code budget worksheet: the figures a manager fills in before and after a pilot

A budget for Claude Code has two parts: a reference figure to start from, and a measured figure to replace it with. Anthropic supplies the first. Its documentation on managing costs, read on 4 October 2026, states an average cost across its enterprise deployments of $13 per developer per active day and $150 to $250 per developer per month, with costs below $30 per active day for 90 percent of users, and in the same paragraph says per-developer costs vary widely and advises a small pilot group to establish a baseline before a wider rollout [1]. The worksheet on this page is built around that advice. It is the last page of the guide to Claude Code costs for teams, and it links to the page that explains each line.

The unit behind every figure is the token, a piece of text the model reads or writes, which Anthropic's pricing page puts at 4 characters or 0.75 of an English word [3]. A developer's cost is the cost of the tokens their sessions use. On a Claude for Teams or Enterprise plan each developer has a seat, a paid place on the plan, with an allowance, an amount of usage included in the seat's price that resets on a rolling five-hour window and a weekly window; Anthropic says usage inside that allowance is not metered in dollars, and dollar figures appear only for usage credits, the paid usage a member draws on after the allowance once an administrator turns them on [1]. On the Claude Console, Anthropic's account for programs that call its models, and on a cloud provider, usage is billed per token [1]. The worksheet asks for the setup first because every later line depends on it.

Anthropic's figures multiplied out, and why the products are a reference only

The table below is this page's own arithmetic. Each cell multiplies one of Anthropic's three figures by a headcount of 10, 50 or 200. The figures are Anthropic's; the multiplication is Reveneau's; and the products are a forecast for no team, for four reasons given immediately after the table.

Developers	Anthropic's $13 per developer per active day, multiplied	Anthropic's $150 to $250 per developer per month, multiplied	Anthropic's $30 per active day, the figure 90 percent of users stay below, multiplied
10	$130 per active day	$1,500 to $2,500 per month	$300 per active day
50	$650 per active day	$7,500 to $12,500 per month	$1,500 per active day
200	$2,600 per active day	$30,000 to $50,000 per month	$6,000 per active day

First, the $13 and the $150 to $250 are averages Anthropic states for other companies' deployments, and Anthropic's own page says costs vary widely with model selection, codebase size and usage patterns such as running several instances or automation [1]. Second, Anthropic gives no definition of an active day, and does not say how many active days a month the monthly figure assumes, so a product of $13 and a headcount assumes every developer is active on the same day. Third, the $30 figure is the level that 90 percent of users stay below, so the last column multiplies a near-upper figure by every developer, which describes a team in which everyone is at the 90th percentile at once. Fourth, on a Teams or Enterprise plan the dollar figure applies to usage credits only, because usage inside the seat allowance is not metered in dollars [1]. The table is a set of reference points to compare a measured figure against. The page on what Claude Code costs per developer explains each of Anthropic's three figures and what moves them.

Part one: the team and the setup

The first lines of the worksheet record the number of developers who will have access and the setup they will use. Anthropic's documentation gives three setups and says the controls differ for each: a Claude for Teams or Enterprise plan, where usage draws from each member's seat allowance; the Claude Console, where usage is billed per token to the organisation; and a cloud provider, Amazon Bedrock, Google Cloud's Agent Platform or Microsoft Foundry, where usage is billed per token to the cloud account [1]. If the organisation mixes sign-in methods, each developer is metered according to the one they authenticated with, so the worksheet has a line for that too [1]. On a seat plan, record the seat tier, Standard or Premium, because Anthropic says the allowance's size depends on it and the allowance is shared with Claude chat and Cowork [1]. The page on subscription seats or pay per token is the page for this decision.

Part two: the pilot and its measured figures

The pilot section records the group size, the start and end dates, and whether the pilot covers at least one full weekly allowance window, which is Reveneau's recommendation since Anthropic advises a pilot without naming a length [1]. The measured section is the most important part of the worksheet. The first line is the total pilot spend, read from the screen Anthropic names for the setup: the spend report in organisation analytics on a Teams or Enterprise plan, which shows estimated spend per user and per model with CSV export, updated daily, and appears once usage credits are turned on; the usage page and the dashboard at platform.claude.com/claude-code on the Claude Console; or the cloud billing console on a cloud provider, because Claude Code sends no metrics from a cloud back to Anthropic [1][2]. The page on where to see Claude Code spend gives every screen.

The second line is the active day count. The analytics dashboard for Teams and Enterprise plans shows daily active users and sessions, and the Console dashboard shows the same on its Activity chart [2]. The third line is the cost per active day: the spend divided by the active days. That division turns the pilot's figure into the same unit as Anthropic's $13, and it is the one figure every other line of the worksheet is built from. A fourth line records the active days per developer per month that the pilot showed, and a fifth multiplies the two to give a cost per developer per month, in the same unit as Anthropic's $150 to $250. A sixth records the share of spend on Opus, from the per-model breakdown, because Anthropic names Opus left as the default model as one of the two usual causes of unexpectedly high spend [1]. Anthropic says the Console dashboard's spend figures are estimates for analytics purposes and the spend report shows estimated spend, so the final line of the section is the invoice total, for reconciliation [1][2]. The page on per-developer usage reporting explains how to get the per-person numbers.

Part three: caps, rate limit and reporting

The caps section asks where the cap is set for the chosen setup and what it is. On a Teams or Enterprise plan, Anthropic names the seat allowance as the default upper limit and says that to let members continue past it, an administrator turns on usage credits and sets spend limits at the organisation, group or individual member level [1]. On the Claude Console the cap is a workspace spend limit on the Claude Code workspace, a named section of the account created automatically the first time Claude Code authenticates [1]. On a cloud provider it is the cloud's own budget controls, and an organisation that runs a self-hosted gateway has a per-user cap there as well [1]. The page on spend limits for an organisation, a group and a person gives each one.

For a Console setup there is a rate limit line too. A rate limit is the most work the organisation may send to the model in one minute, and Anthropic's table recommends a tokens-per-minute figure per user for each team size, with an example of 200 users at 20,000 each, or 4 million in total [1]. Record the row your head count falls in, the organisation total you requested, and whether a workspace rate limit was set on the Claude Code workspace to protect other workloads [1]. The page on rate limits per user by team size reproduces the table. The reporting lines record which per-user source the organisation uses and who pulls it: the spend report CSV on the Teams plan, the Enterprise Analytics API on the Enterprise plan, the Console dashboard or the Claude Code Analytics API on the Console, and OpenTelemetry export, which Anthropic says works on every setup, on a cloud provider [1].

Part four: policy and the monthly review

The policy lines record the decisions an administrator makes once. Managed settings are settings the organisation deploys to every developer's machine, which a developer's own settings cannot override. The worksheet asks whether modelPricing, the managed setting that makes Claude Code's own cost figures match a contracted rate, has been deployed, and notes that it requires Claude Code version 2.1.242 or later [1]; whether the default model has been decided; and whether autoContinueAtUsageLimit, the setting that decides if a session waits and continues on its own after a usage limit, has been set for the fleet [1]. The page on settings an administrator can enforce covers each key.

The review lines name the person who reviews the worksheet each month, the review date, the limit messages developers reported that month, and the Claude Code version in use. Anthropic says Claude Code regularly receives updates that may change how features work, including cost reporting, and names claude --version as the check [1]. The page on what each limit message means is the reference for the messages line.

The worksheet

Section	Line	Your entry	Where the figure comes from
Team	Developers with access		Your own count
Team	Setup: Teams or Enterprise seats, Claude Console, or cloud provider (name which)		Anthropic's three setups [1]
Team	Seat tier, Standard or Premium, if on seats		Anthropic: allowance size depends on tier [1]
Team	Mixed sign-in methods, yes or no		Anthropic: each developer metered by their method [1]
Pilot	Pilot group size		Your choice; Anthropic advises a small group [1]
Pilot	Start date and end date		Your choice
Pilot	Covers one full weekly window, yes or no		Reveneau's recommendation
Measured	Total pilot spend		Spend report, Console usage page and dashboard, or cloud billing console [1][2]
Measured	Total active days across the group		Analytics dashboard or Console Activity chart [2]
Measured	Cost per active day (spend divided by active days)		Your arithmetic; compare with Anthropic's $13 and $30 [1]
Measured	Active days per developer per month		Your own count; Anthropic gives none [1]
Measured	Cost per developer per month (cost per active day times active days)		Your arithmetic; compare with Anthropic's $150 to $250 [1]
Measured	Share of spend on Opus		Spend report per model, or Console dashboard [1][2]
Measured	Invoice total for the pilot period		Billing page; dashboard figures are estimates [1][2]
Caps	Where the cap is set for this setup		Admin settings, Claude Code workspace, cloud budget, or gateway [1]
Caps	Organisation limit		Organization settings > Usage, once usage credits are on [1]
Caps	Group and member limits		Same screen [1]
Caps	Workspace spend limit on the Claude Code workspace		Claude Console [1]
Caps	Cloud budget or gateway cap		Cloud billing console or gateway [1]
Rate limit (Console)	Row of Anthropic's table for your head count		Anthropic's rate limit table [1]
Rate limit (Console)	Organisation total requested		Your arithmetic on Anthropic's per-user figure [1]
Rate limit (Console)	Workspace rate limit set, yes or no		Claude Code workspace Limits page [1]
Reporting	Per-user source in use		Spend report CSV, Enterprise Analytics API, Console dashboard, Claude Code Analytics API, or OpenTelemetry [1]
Reporting	Who pulls it and how often		Your choice; the spend report is updated daily [1]
Policy	modelPricing deployed, yes or no (needs version 2.1.242 or later)		Managed settings [1]
Policy	Default model decided		Managed settings or admin console [1]
Policy	autoContinueAtUsageLimit decided		Managed settings [1]
Review	Reviewer		Reveneau's recommendation: the person who can change the cap
Review	Monthly review date		Your choice
Review	Limit messages reported this month		Developers
Review	Claude Code version in use		claude --version [1]

How to fill it in, in order

Reveneau's recommendation is to fill the worksheet in two passes. Before the pilot, complete the team, pilot and policy sections, write Anthropic's three figures beside the blank measured lines, and leave the measured lines blank; a guessed active day count produces a guessed budget. After the pilot, fill the measured section from the screens named above, reconcile against the invoice, and only then set the caps, because the caps should come from the measured cost per active day multiplied by the developers and active days you expect, with Anthropic's $30 figure as the point above which the model default and session habits are the first things to check [1]. Keep the sheet with the budget and bring it to each monthly review. The sister guides on how to reduce Claude Code token usage and how the Claude Code prompt cache works are for the developers whose figures the sheet records.

Reveneau is an AI software development consultancy. All of its code is written by AI, and every change must pass an eval suite, a set of automated tests written from the specification, before release, so token use is a running cost of every Reveneau build. Reveneau is independent of Anthropic, and every figure on this page is either Anthropic's own statement about its own product or this page's labelled arithmetic on those statements. If you want help running a pilot and turning its figures into a budget, contact Reveneau.