Actual spend and what Lineman saved, per hour across a day
This repository is not on the authorised list.
Spend ran far above the recent baseline for this window.
This spend landed outside the configured working hours.
Spend has passed the budget set for this period.
Projected from current run-rate vs budget
At this run-rate you will exceed the monthly budget in ~4 days.
Unit economics and Lineman savings
Free plan, no credit card. 30-second install.
Connect the Reporting connector once, then anyone who watches the numbers (finance, managers, developers) can ask Claude about your usage, savings, and budgets in plain English. Straight from Claude Desktop or Claude Code, with no dashboard to learn.
Loaded tools · used Lineman Reporting
Here is your Lineman spend for last month:
Want me to break it down by repo, developer, or cost centre?
Lineman sits between your tools and Claude. It shrinks the heavy stuff and labels the spend, without changing a line of how your team works.
Hooks catch every Read, Bash, Grep and build output before it ever reaches Claude.
A secondary LLM rewrites the noisy bulk into a tight summary: the same signal, a fraction of the tokens.
Every call is tagged from your git context: developer, repo, branch, ticket and model. No manual work.
Claude reads the same files and the same build results. You are simply billed for fewer tokens.
Lineman watches every 15-minute bucket and flags the spend that matters: an unauthorised repository, work outside hours, a budget breach, or a sudden spike, each with the developer, repo and model that caused it already attached.
This repository is not on the authorised list.
Spend ran far above the recent baseline for this window.
This spend landed outside the configured working hours.
Spend has passed the budget set for this period.
A developer just works. Lineman watches the calls, reads the git context, and resolves the whole session to one attributed figure. Zero manual tagging.
Shipping in Claude Code today. Cursor, Cline, Continue and Windsurf all speak the same open protocol, so Lineman travels with you.
No proprietary lock-in. Your editor, your workflow, your keys.
Drop Lineman into Claude Code and start cutting your token bill on the very next command. No prompts to rewrite, no workflow change.
Lineman compresses your spend, explains it, governs it, and keeps your sessions going. Pick a job to see how.
Lineman trims large prompts, file reads and build output before they reach Claude, with no loss of meaning. You send fewer tokens and pay less.
Lineman trims large prompts, file reads and build output before they reach Claude, with no loss of meaning. You send fewer tokens and pay less.
Every Claude Code call is tracked and attributed to a developer, repo and ticket in real time, with no manual tagging.
Lineman flags unauthorised repos, out-of-hours work and budget breaches, each with the developer, repo and model that caused it already attached.
This repository is not on the authorised list.
Lineman summarises your session before Claude's native compaction, so long sessions hold on to far more of what matters.
You are handing Lineman the most sensitive thing you own: your source. So the safe choice is the default. Compression happens in memory, your code never trains a model, and every dollar of Claude Code spend is logged and attributable.
ISO 27001 + ISO 27701 accreditation in progress (SOC 2 to follow). Goo Holdings Ltd (England and Wales), trading as Lineman.
The questions engineering leaders and developers raise most, answered up front.
Yes. The Free plan needs no credit card and gives you the cost and spend reporting dashboard, tracking up to $1,500 per month of AI agent spend across your repos, tickets, and models. Paid plans add Lineman's token-saving tool (40%+ fewer tokens), team seats, and uncapped reporting. You can start a 14-day trial of a paid plan at any time.
Lineman never summarises blindly, every compression is task-aware: the secondary model sees what your model is working on and only drops content demonstrably unrelated to it. And it stays risk-free, because Claude keeps control. Whenever it judges a summary too thin for the task, it falls straight through to the full, verbatim output, so you never lose anything you actually needed. The false-drop rate sits at 0.4% across our benchmark suite.
Tool output is processed in memory only and discarded as soon as the response is returned. Nothing is logged, trained on, or persisted unless you opt in to telemetry.
Today, Lineman works with Claude Code. Cursor, Cline, Continue and Windsurf are coming soon. Because they all speak the open MCP standard, support is mostly a matter of testing, and we only ship a client once we test it daily, which right now means Claude Code.
Only the input tokens that pass through Lineman count toward your monthly limit, Basic 100M / month, Pro 440M / month. We never count tokens we drop, and we never count the primary model's response tokens. You can watch the meter live in your dashboard.
Compression keeps working, we never block your dev loop, but additional tokens are billed at a small per-million rate. If you're consistently over, switching plans saves money; the dashboard tells you when.
One config line, works with Claude Code today. Most users see their first compressed call within 30 seconds.