See, forecast and verify your AI coding spend.

Claude Code, Cursor, Codex and Copilot in one view.
Spend by team and repository, a month-end forecast, and each vendor invoice checked against your seats and usage.

Book a demo See pricing

app.lodeline.ai/dashboard

app.lodeline.ai/dashboardSample data

Overview

Last 30 days

Annual budget pace

Run-rate $323,280 a year (last 90 days), 108% of the $300,000 you declared.

AI spend · Last 30 days

$26,940

+9% vs previous period

Cost per developer

$449

60 developers

Active tools

Claude Code, Cursor +2

Open opportunities

$2,180

/mo · from $730 · 3 findings

Cache underuse · Payments

$1,120/mo est. · from $370· of $8,960 spent

Payments reads 11% of its input from cache. A cache read costs about a tenth of fresh input.

Cache underuse · Platform

$640/mo est. · from $210· of $7,480 spent

Platform reads 16% of its input from cache. A cache read costs about a tenth of fresh input.

Cache underuse · Growth

$420/mo est. · from $150· of $5,600 spent

Growth reads 19% of its input from cache. A cache read costs about a tenth of fresh input.

Works with

Claude CodeOTLP push
CodexOTLP push
CursorRead-only API
GitHub CopilotRead-only API

The market

AI coding spend is outrunning the budget.

Three published figures, each with its source.

2028

AI coding costs pass the average developer's salary

Gartner prediction, 24 June 2026

43%

of tech leaders spend more on AI than they planned for 2026

Retool, AI cost management survey, July 2026 (n=101)

98%

of FinOps practitioners now manage AI spend

FinOps Foundation, State of FinOps 2026 (n=1,192)

Forecast

Forecasts that state their own miss.

The month-end total with its likely range.

Sample data

Spend forecast

$28,259expected by the end of October

Likely $26,882 to $29,636Medium confidence

$15,737 spent so far

From your median daily spend over the last 60 days. Past month-end forecasts missed by 11% on average.

Alerts

Alerts when spend moves

When a budget crosses its mark or a day's spend breaks from its usual pattern, the Slack channel you choose gets a message with the figures.

payments-api · daily spend

Sep 24 to Oct 14

eng-spend

Budget and spend alerts for engineering

Sample data

LodelineApp9:02 AM
Budget "Platform" is at 82%: $6,150 of $7,500 (crossed the 80% mark), period from 2026-10-01.
LodelineApp9:41 AM
Spend on repo payments-api for 2026-10-14 was 3.62x above what recent Wednesdays predicted ($1,840 against an expected $508).

Every dollar, traced to the team that spent it.

payments-api · daily spend

Sep 24 to Oct 14

Every dollar, broken down

Spend by tool, team, repository and model, with the month-end forecast beside it.

Sample data

Spend forecast

$28,259expected by the end of October

Likely $26,882 to $29,636Medium confidence

$15,737 spent so far

From your median daily spend over the last 60 days. Past month-end forecasts missed by 11% on average.

Forecasts that state their own miss

The month-end total with its likely range, and 3, 6 and 12 month ranges for planning.

app.lodeline.ai/dashboardSample data

Opportunities

Ranked by monthly savings, measured from your own usage.

Cache underuse · Payments

$1,120/mo est. · from $370· of $8,960 spent

Payments reads 11% of its input from cache. A cache read costs about a tenth of fresh input.

Cache underuse · Platform

$640/mo est. · from $210· of $7,480 spent

Platform reads 16% of its input from cache. A cache read costs about a tenth of fresh input.

Cache underuse · Growth

$420/mo est. · from $150· of $5,600 spent

Growth reads 19% of its input from cache. A cache read costs about a tenth of fresh input.

Waste, with the evidence

Each finding shows what was measured and how much a month it could recover.

Invoice check · September 2026

Sample data

1 of 2 need a look

CursorCheck this

Expected bill $6,180Invoiced $7,040Difference -12%

Cursor billed $7,040.00 but its plan and usage add up to $6,180.00, so $860.00 is not accounted for.

Claude CodeMatches

Expected bill $18,420Invoiced $18,610Difference -1%

Claude Code matches its invoice within 5%.

Invoice checked

Each vendor bill against your seats and billed usage, with differences flagged.

eng-spend

Budget and spend alerts for engineering

Sample data

LodelineApp9:02 AM
Budget "Platform" is at 82%: $6,150 of $7,500 (crossed the 80% mark), period from 2026-10-01.
LodelineApp9:41 AM
Spend on repo payments-api for 2026-10-14 was 3.62x above what recent Wednesdays predicted ($1,840 against an expected $508).

Alerts in Slack

A message with the figures when a budget crosses its mark or a day breaks pattern.

Connections

Sample data

Claude Code
OTLP
Connected
Last synced
2m ago
Volume
~4,120 events a day
Codex
OTLP
Connected
Last synced
6m ago
Volume
~1,310 events a day
Cursor
Admin API
Connected
Last synced
38m ago
Volume
~2,870 events a day
GitHub Copilot
Admin API
Waiting for data

Connected in minutes

Four tools, each with its own status, read-only or write-only.

Live the same day

1
Point your tools at us
An OTLP endpoint and a write-only key: an env block in ~/.claude/settings.json for Claude Code, or a short block in ~/.codex/config.toml for Codex. Cursor and Copilot connect with a read-only admin token instead.
2
Watch the first numbers arrive
Claude Code and Codex within minutes, Cursor and Copilot after the first hourly sync. Nothing to install.
3
Check the invoice
Add each tool's plan and record its bill. We check it against seats and billed usage and flag differences.

Add to ~/.claude/settings.json, merging with any env block already there, then restart Claude Code.

{
  "env": {
    "CLAUDE_CODE_ENABLE_TELEMETRY": "1",
    "OTEL_METRICS_EXPORTER": "otlp",
    "OTEL_EXPORTER_OTLP_PROTOCOL": "http/json",
    "OTEL_EXPORTER_OTLP_ENDPOINT": "https://ingest.lodeline.ai",
    "OTEL_EXPORTER_OTLP_HEADERS": "Authorization=Bearer <your write-only key>",
    "OTEL_EXPORTER_OTLP_METRICS_TEMPORALITY_PREFERENCE": "delta",
    "OTEL_METRICS_INCLUDE_REPOSITORY": "true"
  }
}

Connections

Sample data

Claude Code
OTLP
Connected
Last synced
2m ago
Volume
~4,120 events a day
Codex
OTLP
Connected
Last synced
6m ago
Volume
~1,310 events a day
Cursor
Admin API
Connected
Last synced
38m ago
Volume
~2,870 events a day
GitHub Copilot
Admin API
Waiting for data

Invoice check · September 2026

Sample data

1 of 2 need a look

CursorCheck this

Expected bill $6,180Invoiced $7,040Difference -12%

Cursor billed $7,040.00 but its plan and usage add up to $6,180.00, so $860.00 is not accounted for.

Claude CodeMatches

Expected bill $18,420Invoiced $18,610Difference -1%

Claude Code matches its invoice within 5%.

Pricing

One fee, set by the budget you declare.

Fixed fee for a 3, 6 or 12-month term; all features included. The factors we measure do not affect the amount you will pay.

Included on every plan

All four tools connected
Spend, forecast and team dashboards
Invoice reconciliation against your vendor bills
Waste detection with evidence
Budget and spend deviation alerts in Slack
Model tier comparison
Unlimited budgets and dashboard users
Email and Slack digest

What is your annual AI coding spend?

Start: up to $200,000
Business: $200,000 to $750,000
Enterprise: above $750,000

Book a demo for a quote

Estimate

How much could you recover?

Developers using AI coding tools60

5300

Per-token AI spend per developer, per month$450

$40$2,000

Your per-token spend

$27,000

/mo

Cache underuseup to $5,265/mo

Recoverableup to 20% of per-token spend

up to $63,180/ year

up to $5,265 a month, at 60 developers.

See it on your own data

Token counts and costs. Nothing else.

What we receive

Token counts and cost, per model
Tool, version and repository name
Session time and a random session ID
Counts of lines changed, never the lines
A salted one-way hash of each tool account

What we never receive

Source code or file contents
Prompts or completions
Keystrokes
Developer names or email addresses from your tools
A score or ranking of any person

No individual evaluated

Built so that no person is ever scored, the use the EU AI Act treats as high-risk at work.

How this holds

The EU AI Act treats systems that evaluate individual workers as high-risk. Lodeline is designed to hold no individual to evaluate: identity is never assembled, only a per-workspace salted hash used for de-duplication, and every figure is a team or repository aggregate.

EU AI Act, Annex III §4(b) · obligations apply 2 Dec 2027 per Reg. (EU) 2026/1744

Isolated per workspace

The database refuses to return another workspace's rows, whatever the query.

How this holds

Each workspace's rows carry its id, and the database itself filters every query by it. Hosting for a specific jurisdiction is agreed per contract.

Row-level security on every table

Yours to export or delete

Every table as CSV, or the whole workspace deleted, from Settings.

How this holds

Export and deletion are self-service: a deletion can be undone for 30 days, and backups keep the data up to 14 days after it.

GDPR Art. 17 erasure and Art. 20 portability

Read the security overview Privacy policy

FAQ

Questions

Anything else: sales@lodeline.ai

Which tools do you connect to?

Claude Code, Cursor, Codex and GitHub Copilot. Each takes a few minutes, and none gives us write access.

How does pricing work?

A fixed fee is applied based on the AI coding budget you have specified. This fee does not vary according to the measurements we take.

Do you track individual developers?

No. We never build a profile of a person. Usage carries only a salted one-way hash used to avoid counting it twice, and every figure is a team or repository total.

Who in our company can see the numbers?

Only the people you invite, with owner, admin or member rights. Every figure is a team or repository total.

How do we get started?

Book a demo. After the call we open your workspace, and you connect your tools the same day.

Know the number before the invoice arrives.

Book a demo and connect your tools the same day.

Book a demo See pricing