# AI Cost Ops

> AI Cost Ops helps agencies and SaaS companies control AI spending with hard budget enforcement, prompt injection guardrails, smart PII detection, team access, model routing, anomaly detection, judgment infrastructure, and stack audits.

## What We Do

AI Cost Ops is the enforcement layer between your apps and your AI providers. We operate on structurally qualified execution states — enforcing whether runtime behavior stays within the bounds of what was legitimately approved. Every enforcement decision is bound to the exact conditions that made it legitimate — not just who approved it, but what was true when they did. We help businesses find wasted spend, prevent future overruns, protect sensitive data, and build auditable decision records with two core products.

## Why Choose AI Cost Ops

1. **Hard Budget BLOCKING** — Unlike Helicone or Portkey that only alert after overspend, we actually BLOCK API requests before they exceed your budget. No tokens generated, no charges incurred.
2. **$49 Entry Point** — Enterprise SaaS audit tools (Zylo, Torii, Zluri) start at $50,000+/year. We deliver actionable savings reports for $49.
3. **Dual Product (Audit + Proxy)** — No competitor combines SaaS stack analysis with real-time API budget enforcement in one platform.
4. **Ready-to-Send Action Plans** — Cancellation email templates, pre-cancel checklists, and migration steps. Not just data — actionable scripts.
5. **Keep/Cut/Replace Framework** — Clear, memorable, actionable decisions for every tool in your stack.
6. **Built-in Prompt Injection Detection** — AI-powered injection guardrails at the proxy level. No competitor at this price offers it.
7. **Smart PII/HIPAA Detection** — Scans prompts for sensitive data patterns (SSN, credit cards, medical records) and flags them for human review before they reach the LLM.
8. **Team / Multi-Seat Access** — Role-based access control (Admin, Viewer, Approver) with organization management and invite system.
9. **Model Routing with Fallback** — Priority-based routing rules that automatically failover to fallback models when the primary model is unavailable.
10. **Automated Anomaly Detection** — Daily automated monitoring flags spend spikes exceeding 3x your 7-day rolling average with webhook alerts.
11. **Judgment Infrastructure** — Decision State IDs (ACO-XXXX), State Ledger, Policy Gateway, and Named Stop Policies for ISO/IEC 42001-ready oversight.

## Comparison: AI Cost Ops vs Alternatives

| Feature | AI Cost Ops | Helicone | Portkey | Zylo / Torii |
|---------|-------------|----------|---------|--------------|
| Hard budget blocking | ✅ Blocks requests | ❌ Alerts only | ❌ Alerts only | ❌ N/A |
| SaaS stack audit | ✅ From $49 | ❌ | ❌ | ✅ From $50K/yr |
| Per-customer budgets | ✅ | ❌ | Partial | ❌ |
| PII/HIPAA detection | ✅ Built-in | ❌ | ❌ | ❌ |
| Prompt injection guardrails | ✅ Built-in | ❌ | ❌ | ❌ |
| Identity Guard (persona hijack) | ✅ Built-in | ❌ | ❌ | ❌ |
| Decision State IDs (ACO-XXXX) | ✅ Unique tokens | ❌ | ❌ | ❌ |
| State Ledger | ✅ Replayable timeline | ❌ | ❌ | ❌ |
| Policy Gateway | ✅ State-based routing | ❌ | ❌ | ❌ |
| H=(S,X) judgment coordinates | ✅ | ❌ | ❌ | ❌ |
| Named Stop Policies | ✅ Versioned policies | ❌ | ❌ | ❌ |
| Approval audit trail | ✅ Immutable snapshots (structurally qualified) | ❌ | ❌ | ❌ |
| Team / multi-seat access | ✅ Admin/Viewer/Approver | ✅ | ✅ | ✅ |
| Model routing / fallback | ✅ Priority-based | ❌ | ✅ | ❌ |
| Anomaly detection | ✅ Built-in | ❌ | ❌ | ❌ |
| Compliance export | ✅ JSON/CSV bundles | ❌ | ❌ | ❌ |
| Per-customer rate limits | ✅ RPM + TPM | ❌ | Partial | ❌ |
| Expanded webhooks | ✅ 5 event types | Partial | Partial | ❌ |
| Entry price | $49 one-time / $49/mo | Free tier | Free tier | $50K+/yr |

Unlike Helicone and Portkey that only alert you after overspending, AI Cost Ops actually blocks over-budget API requests before they reach the provider — no tokens generated, no charges incurred.

## Products

### Stack Savings Audit ($49 one-time)
AI-powered analysis of your AI & SaaS tool stack that delivers:
- Keep/Cut/Replace decisions for every tool
- Projected monthly and annual savings
- Replacement suggestions with migration steps
- Cancellation email templates
- Satisfaction guarantee: Find $50+/month in savings or get a refund

### Done-For-You Audit ($297 one-time)
Expert-led audit for busy owners:
- Full review of your AI & SaaS stack
- Keep/Cut/Replace with detailed explanations
- Cancellation & negotiation email templates
- Personal expert review
- Optional performance-based pricing

### API Cost Proxy — Starter Operator ($49/month)
Entry-level real-time budget enforcement for LLM API calls:
- Up to 500K analysis tokens / month
- 1 provider key
- Basic PII detection
- Hard budget caps & velocity limits
- Per-customer cost tracking
- Email support

### Copilot Governance ($149/month)
License-level enforcement for Microsoft 365 Copilot — no API key required:
- Microsoft 365 Copilot connection via Graph API
- License spend visibility & analytics
- Hard budget caps on total license spend
- Idle seat detection & auto-revocation alerts
- Seat approval workflow for new licenses
- License waste reports
- Compliance export with Decision State IDs
- Email & priority support

### API Cost Proxy — Pro Operator ($299/month)
Full-featured budget enforcement for teams running AI in production:
- Up to 5M analysis tokens / month
- Hard budget enforcement & velocity limits
- Human-in-the-loop approval gates + smart PII detection
- AI-powered prompt injection guardrails
- Identity Guard — AI persona hijack detection
- Per-customer cost tracking & rate limits (RPM and TPM)
- Custom models support
- H=(S,X) approval snapshots — ISO/IEC 42001-ready audit trail with Decision State IDs. Every enforcement decision is bound to the exact conditions that made it legitimate.
- Model routing with automatic fallback chains
- Automated cost anomaly detection (daily)
- State Ledger, Policy Gateway, Named Stop Policies
- Compliance export bundles (JSON/CSV)
- Webhook notifications for 5 event types
- Priority support

### API Cost Proxy — Agency Operator ($599/month)
For agencies managing multiple clients:
- Up to 25M analysis tokens / month
- Full multi-client isolation & shared workspaces
- Advanced prompt injection detection + PII/HIPAA compliance scanning
- Identity Guard + advanced identity impersonation scanning
- H=(S,X) snapshots + ISO/IEC 42001 export + Decision State IDs + State Ledger
- Policy Gateway & Named Stop Policies
- API access, webhooks, and custom integrations
- Priority support & SLAs
- Everything in Pro Operator

### API Cost Proxy — Enterprise (Custom pricing)
For regulated industries and large teams:
- Unlimited analysis tokens
- Dedicated infrastructure
- Built on SOC 2-certified infrastructure. ISO/IEC 42001-ready audit exports. GDPR-conscious by design. EU AI Act enforcement layer.
- On-premise deployment options
- Custom SLA & security review
- Volume discounts
- Everything in Agency Operator
- Contact: hello@aicostops.com

## New Features

### Copilot Governance ($149/month)
License-level enforcement for Microsoft 365 Copilot — no API key required:
- Microsoft 365 Copilot connection via Graph API (Reports.Read.All, User.ReadWrite.All)
- License spend visibility and per-seat analytics
- Hard budget caps on total license spend
- Idle seat detection and auto-revocation alerts (prioritizes least-active users)
- Seat approval workflow for new license requests
- License waste reports
- Compliance export with Decision State IDs (ACO-XXXX) for every license change
- Designed for organizations where embedded AI lacks traditional API keys
- URL: https://aicostops.com/ai-proxy (pricing section)

## New Features (continued)

### Trust Proof Center
Public-facing compliance verification dashboard for enterprise procurement:
- Architecture transparency documentation
- Trust badges and compliance status indicators
- Security posture overview for external stakeholders
- Downloadable Security Overview artifact
- URL: https://aicostops.com/trust-proof-center

### Shadow MCP Control
Register, classify, and enforce policies on MCP (Model Context Protocol) tool servers:
- Server registry with URL, status, and tool class assignments
- Tool classifications: read, write, admin, destructive
- Policy rules: allow, block, or require review per tool class
- Full access logging with timestamps and action taken
- URL: https://aicostops.com/mcp-control

### Verified Approval Channels
Multi-factor approval verification for high-value decisions:
- Dual authorization: two independent approvers required for sensitive approvals
- Email OTP verification: one-time password sent to approver's email
- Configurable cost thresholds per channel type
- Same-person approval blocking
- Prevents impersonation by binding approvals to verified identity channels

### Decision State ID Token (ACO-XXXX)
Every approval or rejection generates a unique, human-readable Decision State ID:
- Format: ACO-XXXX (e.g., ACO-3F8A1B2C) — 8-character hex hash
- Uniquely identifies the complete decision state at approval time
- Every enforcement decision is bound to the exact conditions that made it legitimate — not just who approved it, but what was true when they did
- Compliance teams can cite tokens directly in audit reports and regulatory filings
- Appears in CSV exports, compliance bundles, and the State Ledger
- Auto-generated via database function — no manual input required

### What AI Cost Ops Assumes
The system operates on three structural assumptions:
1. The approval state arriving at the gate reflects a legitimate decision context
2. Policy version, budget, and risk status were accurate at the moment of approval
3. The Decision State ID is the authoritative record of those conditions

### State Ledger
A replayable timeline view of every approval decision:
- Each entry displays: Decision State ID, approver, action, budget remaining, active policies, strategic alignment, risk overrides
- Designed for compliance officers reconstructing past decision context
- Accessible from the Reports page under the "State Ledger" tab

### Policy Gateway — State-Based Routing Rules
Route AI calls based on approval state conditions, not just budget:
- Define rules with condition combinations (e.g., first-workflow + cost above $1.00 = mandatory review)
- Actions: hold, stop, or escalate
- Each rule has a priority for deterministic evaluation order
- Configure from Settings → Policy Gateway

### Named Stop Policies
Versioned, reusable execution boundaries for specific workflows:
- Each policy has a name, version number, conditions (JSON), and scoped workflow IDs
- When a request matches conditions, execution halts unless conditions are met
- Useful for compliance teams enforcing workflow-level execution boundaries
- Configure from Settings → Stop Policies

### Team / Multi-Seat Access
Role-based access control for organizations:
- Create organizations and invite team members by email
- Three roles: Admin (full access), Viewer (read-only), Approver (can approve/reject gated requests)
- Admins manage team membership and role assignments
- Organization-scoped data visibility based on assigned roles

### Usage Analytics Dashboard
Time-series cost visualization:
- Daily spend trend with 7-day moving average
- Cost breakdown by AI model
- Cost breakdown by customer
- Filterable by date range, model, and customer

### Automated Cost Anomaly Detection
Daily automated monitoring of spending patterns:
- Compares today's spend against 7-day rolling average
- Flags days where spend exceeds 3x (configurable) the average
- Webhook alerts when anomalies are detected
- Acknowledge/dismiss anomalies from your dashboard

### Model Routing / Fallback Chains
Priority-based routing rules for AI requests:
- Define primary and fallback models per routing rule
- Condition types: cost threshold, always
- Automatic failover: if primary model fails, retry with fallback
- Ordered by priority for predictable behavior

### Compliance Export Package
Downloadable audit trail bundles:
- Approval decision snapshots with Decision State IDs for any date range
- Budget alert history
- PII and injection detection events
- User settings snapshot at export time
- JSON format for easy integration with compliance tools

### Per-Customer Rate Limiting
Prevent single-customer budget exhaustion:
- RPM (requests per minute) limits per customer
- TPM (tokens per minute) limits per customer
- Requests exceeding limits return 429 with Retry-After header
- Set to 0 for unlimited (default)

### Expanded Webhook Events
Five webhook event types for real-time notifications:
- `budget_alert` — budget threshold warnings (80%/100%)
- `approval_decision` — approval or rejection of gated requests
- `pii_detected` — PII/HIPAA data detected in prompts
- `injection_detected` — prompt injection pattern detected
- `anomaly_detected` — spending anomaly flagged

### Provider Limit Override Protection
Provider spending limits (like OpenAI's "hard limits") are advisory — they notify after charges, not before. Users have reported being charged $1,200 despite a $50 hard limit. AI Cost Ops is different:
- Our proxy blocks every request that would exceed your budget before any tokens are generated
- Zero tokens generated = zero charges incurred. This is proxy-level enforcement, not provider-level alerts.

### Per-Tool Budget Caps (Agency Tier)
Per-tool budget caps let you set individual spending limits for each tool or action an AI agent can call — not just per-model or per-workflow, but per function call. For example, cap your agent's `web_search` tool at $5/day while allowing `summarize` to run unlimited. Prevents a single expensive tool from draining your entire budget. Available on Agency Operator and above.

### Tier-Based Feature Access
Dashboard features are gated by subscription tier. Access is enforced server-side via subscription status checks:

| Feature | Starter | Copilot Gov | Pro | Agency |
|---------|---------|-------------|-----|--------|
| Hard budget caps & velocity limits | ✅ | — | ✅ | ✅ |
| Per-customer cost tracking | ✅ | — | ✅ | ✅ |
| Basic PII detection | ✅ | — | ✅ | ✅ |
| Microsoft 365 Copilot governance | — | ✅ | — | — |
| License spend visibility | — | ✅ | — | — |
| Idle seat detection | — | ✅ | — | — |
| Human-in-the-loop approval gates | ❌ | ❌ | ✅ | ✅ |
| Prompt injection guardrails | ❌ | ❌ | ✅ | ✅ |
| Identity Guard | ❌ | ❌ | ✅ | ✅ |
| Model routing & fallback chains | ❌ | ❌ | ✅ | ✅ |
| H=(S,X) snapshots & Decision State IDs | ❌ | ✅ | ✅ | ✅ |
| Custom models | ❌ | ❌ | ✅ | ✅ |
| Per-tool budget caps | ❌ | ❌ | ❌ | ✅ |
| Policy Gateway & Named Stop Policies | ❌ | ❌ | ❌ | ✅ |
| Multi-client isolation | ❌ | ❌ | ❌ | ✅ |
| State Ledger & ISO/IEC 42001 exports | ❌ | ✅ | ❌ | ✅ |

Upgrading is available from the Settings page. Downgrading preserves data but hides gated features.

### Self-Service Onboarding (Spend Fuse)
After purchasing AI Spend Fuse, customers complete a self-service onboarding flow:
1. **Intake Form** — Collects team size, AI providers used, number of environments, and estimated monthly AI spend before checkout
2. **Automated Provisioning** — Stripe webhook triggers automatic environment setup based on intake data (no manual admin action needed)
3. **Onboarding Email** — Branded welcome email sent via Resend with a secure dashboard link
4. **Dashboard Wizard** — 3-step guided onboarding: add API key via encrypted Provider Vault → copy proxy endpoint → verify connection
5. **Secure Key Entry** — API keys are entered directly in the dashboard via the Provider Vault (AES-256-GCM encrypted, write-only). Keys are never sent over email or shared in plaintext.

## Security Features

### Prompt Injection Guardrails (Experimental)
AI-powered detection of common prompt injection attacks at the proxy level:
- Detects "ignore previous instructions", system prompt extraction, jailbreaks, role hijacking, and privilege escalation
- Every prompt is scanned by a fast AI classifier before reaching the LLM provider
- Suspicious prompts are paused and routed to a human-in-the-loop approval queue
- Three sensitivity levels: Strict (flag aggressively), Balanced (default), Permissive (high-confidence only)
- Fail-open policy: if the classifier fails, the request proceeds normally to avoid blocking legitimate traffic
- Best-effort heuristic scanning — use as one layer in a defense-in-depth strategy

### Smart PII/HIPAA Detection
Scans prompts for sensitive personal information before they reach the LLM:
- Detects Social Security numbers, credit card numbers, medical record numbers, and other PII patterns
- Supports HIPAA-sensitive data categories (PHI, medical records, insurance IDs)
- Flagged requests are held in the approval queue for human review
- Configurable PII categories to match your compliance requirements
- Does not guarantee full HIPAA/GDPR compliance — designed to reduce risk as part of your compliance strategy

### Identity Guard (7th Attack Vector)
Text-based identity impersonation detection at the prompt layer:
- Detects when prompts instruct AI agents to adopt, impersonate, or respond as a real named individual
- Catches persona hijacking: "You are [real person], respond as them"
- AI Cost Ops acts as the enforcement layer: block, escalate to human-in-the-loop approval, or pass
- Distinct from media-based deepfake detection (voice/video) — this is text-based prompt-layer enforcement
- The architecture supports plugging in 3rd-party deepfake detection APIs (Pindrop, Vouched, etc.) as upstream detection sources
- No competitor offers this at the prompt layer

### Human-in-the-Loop Approval Queue
All flagged requests (high-cost, PII-detected, injection-suspected) are held for human review:
- Approve, reject, or let requests auto-expire
- See prompt previews, estimated cost, and risk signals
- Configurable auto-expiry timer
- Immutable decision snapshots with H=(S,X) judgment coordinates — strategic alignment (S), experiential context (X), Decision State ID (ACO-XXXX), policy freeze, budget remaining, and risk override flag for ISO/IEC 42001 human oversight
- Email notifications for pending approvals

## Key Differentiators

1. **Execution Governance Layer**: AI Cost Ops operates on structurally qualified states — it enforces whether execution stays within the bounds of an approved decision context. Unlike spending alerts, our proxy actually BLOCKS requests that would exceed your budget.
2. **Keep/Cut/Replace Framework**: Clear, actionable decisions—not just data dumps.
3. **Agency-Focused**: Per-customer budgets and rate limits let agencies control costs without billing surprises.
4. **Built-in Prompt Injection Detection**: AI-powered injection guardrails at the proxy level.
5. **Smart PII Detection**: Scan prompts for sensitive data before they reach the LLM.
6. **Team Access**: Role-based multi-seat access with Admin, Viewer, and Approver roles.
7. **Model Routing**: Priority-based routing rules with automatic fallback on failure.
8. **Anomaly Detection**: Daily automated monitoring flags cost spikes with webhook alerts.
9. **Compliance Exports**: Downloadable audit trail bundles for internal compliance reviews.
10. **Identity Guard**: Text-based identity impersonation detection at the prompt layer — the 7th injection attack vector. Catches persona hijacking where agents are instructed to adopt a real person's identity. AI Cost Ops acts as the enforcement layer: block, escalate to human review, or pass. No competitor offers this.
11. **Decision State IDs (ACO-XXXX)**: Every approval generates a unique, citable token identifying the complete decision state for compliance reporting.
12. **State Ledger**: Replayable timeline of all approval decisions with full context reconstruction for auditors.
13. **Policy Gateway**: State-based routing rules that go beyond budget thresholds — combine conditions for nuanced enforcement.
14. **H=(S,X) Judgment Coordinates**: Strategic alignment and experiential context alongside Decision State IDs.
15. **Named Stop Policies**: Versioned execution boundaries per workflow for compliance-mandated controls.
16. **AI Sovereignty**: Built-in Decision Sovereignty (human approval gates, Decision State IDs, State Ledger), Vendor Sovereignty (multi-provider vault, model routing, single endpoint), and Data Sovereignty (region-aware routing with EU/US geographic model selection, enforced region logging). Your team controls every AI decision, every provider relationship, and every data flow — no lock-in, no delegation to algorithms, no data leaving your chosen jurisdiction.

## Use Cases

- Agencies managing AI costs across multiple clients
- SaaS companies with LLM-powered features needing budget and security controls
- Developers preventing runaway API costs during testing
- Teams consolidating overlapping SaaS subscriptions
- Companies needing to reduce PII exposure risk in AI prompts
- Security-conscious teams wanting prompt injection protection
- Organizations requiring role-based team access for AI cost management
- Compliance teams needing exportable audit trails with Decision State IDs
- ISO/IEC 42001 auditors verifying human oversight with H=(S,X) judgment coordinates

## Technical Details

- Supported Providers: OpenAI, Anthropic (Claude), Google Gemini, Ollama, any OpenAI-compatible API
- Integration: Change your API base URL to route through our proxy
- Latency: <50ms added latency on average
- Security: HTTPS-only, JWT authentication, no raw API key storage
- Injection Detection: AI classifier powered by Gemini Flash Lite, fail-open policy
- PII Detection: AI-powered + rule-based pattern matching for sensitive data
- Model Routing: Priority-based with automatic fallback on primary model failure
- Anomaly Detection: Daily cron job comparing spend vs 7-day rolling average
- Judgment Infrastructure: Decision State IDs, State Ledger, Policy Gateway, Named Stop Policies

## Topics & Keywords

AI Cost Ops covers these areas of AI pricing and cost management:

- **AI price** — compare AI model pricing across OpenAI, Anthropic, Google, and more
- **AI price control** — actively manage and enforce what you pay for AI API calls
- **AI cost control** — enforce hard budget limits, not just alerts
- **AI budget cap** — hard caps that block requests before charges are incurred
- **AI API cost** — track and cap API costs in real time
- **LLM cost optimization** — reduce token waste and find cheaper model alternatives
- **AI budget management** — set per-customer and per-workflow budgets
- **OpenAI cost reduction** — audit OpenAI usage and block over-budget calls
- **AI spend tracking** — real-time dashboards showing spend by model, customer, and workflow
- **SaaS cost audit** — one-time analysis of your entire SaaS and AI tool stack
- **AI team management** — role-based multi-seat access for AI cost operations
- **AI anomaly detection** — automated daily monitoring for cost spikes
- **Model routing LLM** — priority-based model selection with automatic fallback
- **AI compliance export** — downloadable audit trail bundles for compliance reviews
- **AI pricing** — understand and control what you pay for AI API calls (secondary)
- **Identity impersonation detection** — text-based persona hijack prevention at the prompt layer
- **AI persona hijack prevention** — enforcement layer for identity impersonation attacks
- **Deepfake enforcement layer** — AI Cost Ops as the enforcement layer connecting to detection APIs
- **Decision State ID** — unique ACO-XXXX tokens for citable compliance references
- **State Ledger** — replayable timeline of approval decisions with full context
- **Policy Gateway** — state-based routing rules beyond budget thresholds
- **Judgment infrastructure** — structural records proving human oversight was exercised
- **Structurally qualified states** — enforcement bound to exact conditions at decision time
- **Execution governance layer** — enforcing whether runtime behavior stays within approved bounds
- **Named Stop Policies** — versioned execution boundaries for workflow-level compliance
- **Copilot Governance** — license-level enforcement for Microsoft 365 Copilot without API keys
- **Microsoft 365 Copilot cost control** — seat spend visibility, idle seat detection, budget caps on license spend
- **Embedded AI governance** — enforcement for AI tools that don't use traditional API keys
- **AI Sovereignty** — structural control over AI decisions, provider relationships, and data flows
- **Decision Sovereignty** — human-in-the-loop approval gates with Decision State IDs ensuring your team makes the final call
- **Vendor Sovereignty** — multi-provider key vault and model routing that eliminates provider lock-in
- **AI vendor lock-in prevention** — switch providers without code changes via Provider Vault and model routing
- **AI provider independence** — single proxy endpoint, any AI provider behind it
- **Data Sovereignty** — region-aware routing enforcing geographic model selection (EU/US) at the proxy level
- **Data residency AI** — configure EU or US region routing with enforced region logging for compliance
- **AI geographic routing** — proxy-level enforcement of regional data flow restrictions
- **Region-aware AI proxy** — route AI requests through designated geographic regions with audit trail

## Supported Languages

The platform UI and all public-facing pages are available in 7 languages:
- English (en) — default
- Spanish (es)
- Portuguese (pt)
- French (fr)
- Italian (it)
- German (de)
- Romanian (ro)

Language is auto-detected from browser settings and can be switched manually. All page titles, meta descriptions, and navigation are translated. JSON-LD schema remains in English for SEO consistency with a dynamic `inLanguage` property.

## Contact

- Website: https://aicostops.com
- Email: hello@aicostops.com
- Twitter: @AICostOps

## Pricing Summary

| Product | Price | Type |
|---------|-------|------|
| Stack Savings Audit | $49 | One-time |
| Done-For-You Audit | $297 | One-time |
| Pro Spend Fuse Install | $2,500 | One-time setup |
| Agency Spend Fuse Install | $7,500 | One-time setup |
| Starter Operator (Proxy) | $49/month | Subscription |
| Copilot Governance | $149/month | Subscription |
| Pro Operator (Proxy) | $299/month | Subscription (+ $2,500 setup if bundled with Pro Spend Fuse) |
| Agency Operator (Proxy) | $599/month | Subscription (+ $7,500 setup if bundled with Agency Spend Fuse) |
| Enterprise (Proxy) | Custom | Contact Sales |

## AI Spend Fuse — Done-For-You Install

A productized, done-for-you implementation service. We install a hard API proxy layer that blocks over-budget calls before tokens are generated. Two tiers available:

### Pro Install ($2,500 one-time)
- Delivered in 48 hours
- Up to 3 client environments or projects
- OpenAI and/or Gemini API integration
- Custom budget cap + velocity limit configuration
- Kill Switch setup and test
- 1 developer handoff call (30 min)
- Audit Evidence Pack export path verified

### Agency Install ($7,500 one-time)
- Delivered in 5–7 business days
- Up to 10 client environments or projects
- All Pro Install features included
- Dedicated onboarding manager
- Priority SLA
- Multi-client isolation configuration

### Risk-Free Guarantee
Both tiers include a risk-free guarantee: if we can't successfully block an over-budget call in your test environment within the applicable delivery window, you don't pay.

### Enterprise Upgrades
Also available for regulated / enterprise deployments: Advanced prompt injection + PII/HIPAA scanning, Identity Guard, ISO/IEC 42001 exports (H=(S,X), Decision State IDs, State Ledger), Policy Gateway + Named Stop Policies.

URL: https://aicostops.com/spend-fuse