Know what your AI agents cost.
Know what to do next.
aibill connects coding-agent work to cost evidence, attribution, and one next action. Today's beta runs privately on your machine; the shared company accountability layer comes next.
Reads the tools you already use
Local by default
Raw transcripts stay on-device.
Source linked
Every number names its origin.
Honest gaps
Missing data stays missing.
One contract
CLI, Glance, and MCP agree.
One bill. Three ways to use it.
Each surface keeps the same evidence rules: source, freshness, billing class, and missing data stay explicit.
Stay in flow
See risk before it interrupts you.
Hover at the menu bar for spend, runway, focus, freshness, and one next action.Current session · estimated API-equivalent value
$4.18
Codex · GPT-5.6agent-finops
ChatGPT Pro · detected locally · API value ≠ added spend
Local estimate · Exhausts ~1hCodex reported · Resets in 2h
Local estimate · Below capCodex reported · Resets Monday
Main focus · 7d · local activity
Refining Glance hover UITask · agent-finops · GlanceView.swiftLocal: Claude Code + Codex · 60 files · nothing uploaded
Updated 12s agoGlance and Ask aibill share one illustrative receipt. Terminal replays the real CLI with deterministic sample data. Estimated value, provider-reported cost, detected plans, and missing limits remain visibly separate.
The accountability gap
Financial accountability starts with four questions.
Today's beta answers only what local and provider evidence can support. Company-wide reconciliation, accepted outcomes, approvals, and ROI stay explicitly next.
What work is driving our AI bill?
01Attribute observed activity and cost evidence by project, model, agent, workspace, user, or client—only where the source exposes it.
Who owns it—and did it produce an accepted outcome?
02Observed ownership is available where supported. Accepted-task and pull-request receipts are the next open contract.
Which subscriptions and provider charges never reach finance?
03Local plan context and optional provider reports are available now. A centralized seat and invoice ledger is Workspace next.
What changed, what needs approval, and did the action work?
04The beta surfaces anomalies and one bounded recommendation. Shared approvals and verified results come with Workspace.
Build the financial accountability system with us.
The planned Workspace will connect agent work to costs, owners, accepted outcomes, budgets, approvals, and verified results—so engineering and finance can act from the same evidence.
Workspace is not launched. Local mode stays free and private.
The short answers
Is API-equivalent value my bill?
No. It is comparison math based on published API rates. aibill keeps that separate from detected subscription context and optional provider-reported cost; final invoices may still include credits, discounts, tax, or adjustments.
What leaves my computer?
The default CLI and Glance run locally with no account or telemetry. Provider credentials are used only when you explicitly connect an official billing API, and MCP results go only to the AI client you invoke under that client’s data policy.
What can I use today?
The local CLI and explicit MCP integration are in public beta. Glance can be built from source for testing; a signed Mac download and the shared Workspace are not launched yet.
Can finance use aibill to prove ROI?
Not from spend evidence alone. The beta establishes cost, activity, attribution, and coverage. Defensible ROI requires reconciled cost, an accepted outcome, and independently evidenced business value; those outcome and company-accountability layers are next.