● Free · no signup · self-report

Your AI agent can drain a month's budget overnight. Put a hard cap on it.

A recursive or retry loop can call a model repeatedly on your key. AgentGuard evaluates hard spend caps and an opt-in circuit breaker inside your process before provider dispatch.

Scan your agent, free → See supported bindings

Your provider key goes directly to your selected provider, not through AgentGuard. Free production use covers up to 10,000 enforcement calls per calendar month.

Why observability is not enough

Helicone, Langfuse, and provider dashboards are good at one thing: telling you what already happened. By the time the alert fires, the tokens are spent and the bill is real. A spend cap is a different job. It has to refuse the call at the moment it would exceed your limit, in the same process, before any money moves. That is a control, not a chart.

Hard caps that block, not warn

Per-call, per-day, and per-month ceilings. A guarded call blocks before provider dispatch when a projected total exceeds a configured block cap.

An opt-in loop circuit breaker

Configure repeated-call, plan-churn, and reasoning-step thresholds. Matching guarded calls block before provider dispatch.

Capability gates

Compare a declared capability with the policy requirement before a guarded provider call. For stronger claims, configure a trusted signed DAG attestation for the selected capability.

Optional signed decision records

Configure Ed25519 keys and a decision store to sign and hash-chain guarded decision metadata. A verifier with the trusted public key can check the chain offline.

Add a cap with a complete policy

Use the exported wrapper for your client shape. Policy runs in your process before dispatch; prompts travel directly to the provider you select and do not traverse AgentGuard infrastructure. The circuit breaker is opt-in under the wrapper's guard configuration.

import Anthropic from '@anthropic-ai/sdk';
import { withSpendGuardAnthropic } from '@agentguard-run/spend';

const scope = { tenantId: 'acme', agentId: 'my-agent' };
const guarded = withSpendGuardAnthropic(new Anthropic(), {
  policy: {
    id: 'daily-agent-cap',
    name: 'Daily agent cap',
    scope,
    caps: [{ amountCents: 2000, window: 'per_day', action: 'block' }],
    mode: 'enforce',
    version: 1,
    effectiveFrom: '2026-07-27T00:00:00Z',
  },
  scope,
  config: {
    circuitBreaker: { enabled: true, repeatThreshold: 3, windowSeconds: 300 },
  },
});
// A call that would take daily spend above $20 throws before provider dispatch.

Using LangChain, OpenRouter, Claude Code, or the Vercel AI SDK? See the framework integrations. Want us to handle AI access with no key to manage? That is Managed, $99/mo.

Complete the self-reported checklist

The free scan is a six-field, self-reported governance-posture checklist. It returns checklist recommendations and an illustrative estimate based on the values you enter. It does not inspect your agent. No signup or installation is required.

Scan your agent, free →