Employees in every department now use AI tools, and each department uses different ones. Engineers work in Claude Code and Codex, marketing drafts campaigns in Claude Desktop, sales researches accounts before calls, and support summarizes long ticket threads. Each of those tools sends requests to a model, and each request costs money. As that spend grows into a real line item, a company needs to know exactly where it goes: which people, which tools, which models, which vendors and which regions, and whether each team is on track for the month. That takes attribution that is precise to the request and accurate to the cent.
Attributing Spend With Multi-Router
Multi-Router tracks every person and every tool they use independently. Alice in Claude Code and Alice in Codex are two separate lines of spend, and inside each one every model she uses is tracked on its own. Every person belongs to an organization, such as Engineering or Marketing, so the same spend also adds up by department.
Each request is priced the moment it completes, from the exact model, vendor, variant and region that served it, at the price in effect at that time. Input, output, cache read and cache write tokens are each priced at their own rate, so the cost reflects what the vendor charges, and the savings from prompt caching show up per person and per tool. Each request also records the region that served it, so cost and data residency are visible side by side.
The cost is counted against the person and the tool, and it appears within seconds:
| Where | What it shows |
|---|---|
| The person's page | Spend this period across all their tools, and each tool's spend against its budget. |
| The Portal | The person's own spend by day, their pace, and their spend on each model. |
| Traffic | Every request with its person, model, vendor, region, tokens and cost. |
| Budgets | Spend, status and pace for each budget, with spend by person and by model. |
Alice Chen
activeMemberalice@example.com
OperatorManaged agentsPortal
Access
EditAI agents
Policies
Activity
Lifecycle events, newest page first.
Identity
- alice@example.com
- Account role
- Member
- Sign-in
- SSO linked
- Person ID
- per_2Fk7Ms4QbY
- User ID
- usr_8Wc2Hn5RzE
- Created
- Nov 7, 2026
- Updated
- Nov 14, 2026
Budget
The Employee Portal
The Portal is where each person sets up their AI tools. A person signs in with their work email, sees the tools their policy grants, and sets each one up with a key of their own, the settings to paste into the tool, and a check that confirms it works. There is nothing for IT to configure by hand on anyone's laptop, and every request from then on is attributed to that person and that tool.
Each tool has its own page in the Portal. It lists the models the person can use, with their spend and cap on each, and shows their spend against their budget, their pace, and the date they will reach their budget at that pace. Models the policy does not allow are refused. A person who needs more can request additional budget from the same page, and an approver receives an email and approves or denies it in the dashboard. A person can also issue themselves a new key at any time, and their spend and budget carry over to it.
Your AI agents
2 set up · budgets reset Dec 1
| AI agent | This month | At this pace | Status | |
|---|---|---|---|---|
Claude Code 3 models approved · key sk-mr-7Qx2… | $142.60 of $20071% | cap in 8 days $7.51/day | nearing cap | Details |
Codex 2 models approved · key sk-mr-3Fp8… | $38.40 of $10038% | won’t reach cap before reset $2.02/day | on track | Details |
Fine-Grained Budgets
Each person has their own budget for each tool. Alice can have $200 a month for Claude Code and $100 a month for Codex, and each budget alerts, pauses and resets on its own, so reaching her Codex budget leaves Claude Code working.
Inside each tool budget, a model can have its own limit. On a $200 Claude Code budget, Claude Opus 5 can be limited to $120, so a person uses the most capable model for the work that needs it and keeps working on the other models once Opus reaches its limit.
Budgets also cover more than one person:
| Budget covers | Example |
|---|---|
| A team or department | Engineering's total across every person and tool, $20,000 a month. |
| A chosen set of tools | A pilot group's tools, each with its own limit. |
| A tag | Everything carrying the tag, including tools set up later, with a total for the group and a limit for each member. |
Budgets stack. A department budget, a person's budget for a tool and a model limit all apply together, and the first one reached takes effect, so a department can have a monthly ceiling while each person keeps their own budget inside it.
Rules decide what happens as spend grows: an email at 80%, an alert when spend is projected to reach the limit within the next hour, and a pause at 100% until the period resets. Alerts also go to the systems the company already uses, by webhook to Slack, an on-call service or a finance tool, and every webhook is signed so the receiver can verify it came from Multi-Router. An admin can raise any limit for the current period when a launch or a deadline needs it.
The Budgets page shows every budget's spend, status and pace, and each budget's usage shows spend over time, by person and by model. Traffic exports to CSV for chargeback to cost centers.
All budgets
| Name | Scope | Period | Spend / effective limit | Used | Status | Updated | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Standard Engineering · Claude CodePolicy1 request bdg_2Lk5Pz8NcV | Tag · Standard Engineering | Monthly | $5,840/$200 per person | — | No cap | 3h ago | ||||||||||||||
Standard Engineering · Claude Codebdg_2Lk5Pz8NcVTag · Standard EngineeringMonthly Detailed usage → Managed by policy: limits and rules are set by the policy. Budget Overrides and approvers can still be set here.Open policy Spent $5,840 Effective limit — Remaining — Per-person limit $200base Period Nov 1 – Nov 30resets Dec 1 Pending requests 1 People over cap 0 of 34 Spend per day · daily · Nov 1 – 30 Spend $0$200$400$600 Nov 1 Nov 8 Nov 15 Nov 22 Nov 29 Request approvers Requests onEmailed when someone asks for a higher limit in the Portal. The Portal hides its request button while this list is empty. dana.okafor@example.com name@example.com Add1 of 10 Budget Overrides · this period Nov 1 – Nov 30 · UTC. Each override replaces one limit until the period ends.
Pending requests 1 pendingApproving raises this person's limit for this period. Adjust the amount before approving. Priya ShahClaude Code Nov 19, 09:42 UTCLimit $200 → requested $300 (+50%) I am running the payments service migration this week and I am at $196.10 of $200 with 11 days left in the month. Approve at$300+50% from $200 DenyApprove | ||||||||||||||||||||
Standard Engineering · CodexPolicy bdg_5Wc7Hq3MzA | Tag · Standard Engineering | Monthly | $1,215/$100 per person | — | No cap | 3h ago | ||||||||||||||
Data Science pilot bdg_9Bv3Rt6YfJ | Routers · 4 | Monthly | $2,960/$4,000 | 74% | On track | 4d ago | ||||||||||||||
Engineering bdg_4Hn8Ws1KdT | Org | Monthly | $12,480/$20,000 | 62% | On track | 9d ago | ||||||||||||||
Policies
All of this is set once, in a tools policy, for a group of people. A policy sets the tools a group can use, the models each tool allows, the budget per person for each tool, the limits for each model, and the rules that act on them.
| Policy | Tool | Per person | Model limits | Rules |
|---|---|---|---|---|
| Standard Engineering | Claude Code | $200 a month | Claude Opus 5 at $120 | Email the team lead at 80%, pause at 100% |
| Standard Engineering | Codex | $100 a month | Email the team lead at 80%, pause at 100% | |
| Go-to-Market | Claude Desktop | $40 a month | Email the team lead at 80%, pause at 100% |
Everyone on a policy gets their own budgets the moment they set up a tool, and one edit to the policy applies to all of them.
| Serve | Model | Version | Wire id | Price / 1M in · out |
|---|---|---|---|---|
Claude Fable 5 anthropic/claude-fable-5 | 5 | claude-fable-5 | $10.00 / $50.00 | |
Claude Fable 5.1 anthropic/claude-fable-5.1 | 5.1 | claude-fable-5-1 | $10.00 / $50.00 | |
Claude Haiku 4.5 anthropic/claude-haiku-4.5 | 4.5 | claude-haiku-4-5 | $1.00 / $5.00 | |
Claude Opus 4.5 anthropic/claude-opus-4.5 | 4.5 | claude-opus-4-5 | $5.00 / $25.00 | |
Claude Opus 4.6 anthropic/claude-opus-4.6 | 4.6 | claude-opus-4-6 | $5.00 / $25.00 | |
Claude Opus 4.7 anthropic/claude-opus-4.7 | 4.7 | claude-opus-4-7 | $5.00 / $25.00 | |
Claude Opus 4.8 anthropic/claude-opus-4.8 | 4.8 | claude-opus-4-8 | $5.00 / $25.00 | |
Claude Opus 5 anthropic/claude-opus-5 | 5 | claude-opus-5 | $5.00 / $25.00 | |
Claude Opus 5.5 anthropic/claude-opus-5.5 | 5.5 | claude-opus-5-5 | $4.00 / $20.00 | |
Claude Sonnet 4.5 anthropic/claude-sonnet-4.5 | 4.5 | claude-sonnet-4-5 | $3.00 / $15.00 | |
Claude Sonnet 4.6 anthropic/claude-sonnet-4.6 | 4.6 | claude-sonnet-4-6 | $3.00 / $15.00 | |
Claude Sonnet 5 anthropic/claude-sonnet-5 | 5 | claude-sonnet-5 | $2.00 / $10.00 |
Each person on this policy gets $200 for Claude Code. Resets on the 1st, 00:00 UTC.
A model cap sits inside each person's total and cannot exceed it.
A rule watches the per-person cap, every model cap, or one model cap, and fires when spend reaches a line. Email and webhooks send once per period; block answers 429 until reset.
- Organization
- Engineering
- Created
- Jun 12, 2026
- Created by
- Priya Shahpriya@example.com
- Revision
- 3 · Sep 2, 2026
Save Claude Code before saving the policy.
- Revision 3 · Claude Code
Sep 2, 2026 · Priya Shah (priya@example.com) - Revision 2 · Claude Code
Jul 21, 2026 · Priya Shah (priya@example.com) - Policy created
Jun 12, 2026 · Priya Shah (priya@example.com)
With directory sync, attribution follows the company's org chart. Directory groups assign each person's organization and policy, so when a person moves to another team in the identity provider, their policy and their spend move with them. When a person leaves, their tools stop and the record of their spend stays. Part 3 of Rolling Out AI Company-Wide covers directory sync in detail, and the Budgets & Spend Control docs cover every budget setting.