Product Spend Guard
// Spend Guard · Budget that manages itself

Spend the operator caps and reallocates toward outcomes.

Instead of a team watching dashboards and cutting budgets by hand, the operator hard-stops the call before the provider and moves budget toward what works — on the record. Cut AI spend and the cost of running it, provably.

Every spend number meters on exactly one billable receipt per request, so routing, retries and tool fan-out never double-count.

Book a demo See it in the console →
// What it operates

It enforces the budget in the request and proves the number in the report

Instead of finance reconciling spend after the fact, the operator holds budgets per team and per agent and enforces them with hard-stops in the data path: when a budget is exhausted, it stops the request before the provider call, marks the outcome blocked — and still writes a receipt.

It marks exactly one receipt per request billable. Routing children, cache lookups, guardrail checks and tool calls all chain, but the operator meters none of them — so a single request that fans out across five providers counts once, and the spend number is one you can verify.

// What it gives you
  • Per-team / per-agent budgets. The operator scopes limits to the unit you manage.
  • Inline hard-stops. An exhausted budget stops the call before it costs another cent.
  • One billable receipt per request. Fan-out chains but never double-counts.
// On the record

Every dollar it caps resolves to a receipt

Instead of asserting the savings, the operator records them: every dollar of attributed cost traces to a billable receipt in the chain, so finance and audit read from one source they can both verify.

The operator meters cost on the billable receipt

The cost_usd, tokens_in and tokens_out fields live in the canonical receipt, and exactly one parent per request carries billable=true. Recompute the leaf yourself to confirm the metered amount.

This is the same client-side verification shipped on the Trust portal. SHA-256 runs via Web Crypto. No Agentics call.

// Verify in your browser Anchored
receipt_idc4d8e2f0…4e5f
prev_hash1af67611…4a032
payload_hash2962f6b5…466dd
recomputed—
merkle_root14a24583…c058a
anchor_txblock 420669398 ↗
  • •Recompute leaf
  • •Chain integrity
  • •Fold Merkle proof
  • •Anchor memo == root

SHA-256 runs in your browser via Web Crypto. No Agentics call.

// The mandate

You set the budget. It enforces the hard-stop.

Instead of a person watching the meter, set a monthly budget for a cost center; the operator stops hard the moment it's reached.

{
  "budget": {
    "scope": "cost_center:support",
    "period": "monthly",
    "limit_usd": 5000,
    "on_exhausted": "hard_stop"
  }
}
// over budget → operator blocks pre-call, still receipts it (outcome: blocked)
// metered on exactly one billable=true parent receipt per request

Read the billing model →

// Supervise it

Watch what the operator is capping

Instead of stitching reports together, the cost console shows spend by team and agent against budget, every figure sourced from billable receipts.

acme.agentics.you/console/cost/

Spend Guard console

Spend-against-budget by team and agent, every figure traceable to the billable receipt the operator metered it on.

Open Spend Guard →
// Provable spend

It caps the spend. You verify the number.

Book a demo Read the docs → Verify a receipt →