LLM FinOps

Govern LLM Spend Before It Reaches Provider Bills

Cloptima gives platform and finance teams one control plane for AI gateway policies, bring-your-own-key access, agent attribution, model pricing, and reconciliation-ready reporting.

app.cloptima.ai/llm
AI FinOps · Spend Overview
Last 30 days · all teams
Illustrative
Monthly LLM spend
$42,180
12% vs last month
Tracked events
2.4M
Active virtual keys
18
Budget used
68%
on pace
Spend by model
OpenAI · gpt-4o$16,030
Anthropic · Claude Sonnet$11,390
Google · Gemini 1.5 Pro$7,590
OpenAI · gpt-4o-mini$4,640
Other$2,530
Spend by team
Platform$14,340
Support AI$10,970
Search$8,860
Data$8,010

Govern access before usage becomes spend

Approve providers, models, environments, and policy scopes before applications create material AI costs.

Replace provider-key sprawl with virtual keys

Give each app a governed access path while provider credentials stay protected behind policy.

Apply budget decisions in the request path

Use low-latency checks and durable usage records instead of waiting for delayed provider invoices.

Map model usage to real owners

Attribute spend by team, app, environment, customer, workflow, agent run, and product dimension.

Make agent costs explainable

Track sessions, retries, tool calls, and workflows so autonomous systems stay accountable.

Prepare finance for provider bill review

Contract pricing, commitments, credits, and invoice adjustments in one auditable ledger, previewed before approval.

Price guardrails, not just enable them

Cost-aware detection with a versioned, audited bypass path — plus exact and semantic response caching with observe-before-enforce.

Cut cost without cutting quality

Model downgrades, prompt changes, and cache-enforce decisions run through a dataset-backed eval before they ship, with canary rollout and rollback.

Supported Provider Surfaces and Model Vendors

Govern known model families with canonical policy names while still allowing exact policy strings for fine-tuned, custom, and newly released provider models.

Provider surfaces
16
External vendors
49
Cataloged models
200

Google Vertex AI

Managed Gemini and Model Garden partner models.

66 models
Anthropic Claude · 16Meta Llama · 12Google Gemini · 11Mistral · 7Qwen · 4xAI Grok · 4DeepSeek · 3AI21 Jamba · 2+5 more

AWS Bedrock

Managed foundation models available through Bedrock.

86 models
Mistral · 14Anthropic Claude · 13Meta Llama · 12Qwen · 7Amazon Nova · 5Cohere Command · 4DeepSeek · 4Google Gemma · 4+8 more

OpenAI

Direct OpenAI-compatible provider access.

4 models
OpenAI GPT-4.1 · 2OpenAI GPT-4o · 2

Azure OpenAI

Microsoft Azure OpenAI Service deployments.

4 models
OpenAI GPT-4o · 2OpenAI o1 · 1OpenAI o3 · 1

Anthropic

Direct Anthropic-compatible provider access.

2 models
Anthropic Claude · 2

Google AI (Gemini)

Direct Google AI Studio developer API.

5 models
Google Gemini · 5

Mistral AI

Direct Mistral AI platform models.

3 models
Mistral · 3

Cohere

Direct Cohere Command foundation models.

2 models
Cohere Command · 2

xAI (Grok)

Direct xAI Grok developer platform.

2 models
xAI Grok · 2

Groq

High-speed LPU inference engine for open models.

2 models
Meta Llama · 2

Together AI

Together AI open model server infrastructure.

2 models
DeepSeek · 1Meta Llama · 1

Fireworks AI

Fireworks AI production model platform.

2 models
DeepSeek · 1Meta Llama · 1

OpenRouter

Unified router and model marketplace API.

2 models
Anthropic Claude · 1OpenRouter · 1

DeepSeek

Direct DeepSeek V3 and R1 reasoning APIs.

2 models
DeepSeek · 2

Perplexity AI

Perplexity Sonar search and reasoning APIs.

2 models
Perplexity Sonar · 2

Z.AI

Direct Z.AI GLM model access via api.z.ai.

14 models
Z.AI GLM · 14

Instrument Your Apps With the Cloptima SDKs

Want attribution without routing through the gateway? Drop in a lightweight SDK, keep your own provider client, and send usage with rich, custom attribution. First-class support for JavaScript/TypeScript, Python, and Go.

JavaScript / TypeScript

Node, Deno, Bun
@cloptima/llm-observability View on GitHub

Python

3.9+
cloptima-llm-observability View on GitHub

Go

1.21+
github.com/cloptima/llm-observability-go View on GitHub
install
npm install @cloptima/llm-observability
pip install cloptima-llm-observability
go get github.com/cloptima/llm-observability-go
typescript
import { initFromEnv, extractOpenAIUsage } from "@cloptima/llm-observability";

const cloptima = initFromEnv();

await cloptima.observeCall({
  provider: "openai",
  model: "gpt-4.1-mini",
  call: () => summaryService.generate(prompt),
  extractUsage: extractOpenAIUsage,
  featureId: "summary_generation",
  workflowId: "support_agent",
  // + team, environment, tenantId, costCenter, and custom metadata
});

Telemetry posts to https://api.cloptima.ai/v1/ai/integrations/sdk/events. Missing config falls back to a disabled pass-through, so local dev and tests never break.

Explore the LLM FinOps Platform

Choose the area that matches your rollout: gateway governance, BYOK, budgets, analytics, bill review, unit economics, agents, shadow AI, or private gateway deployment.

AI Gateway

AI Gateway for LLM FinOps

Route model traffic through an OpenAI-compatible gateway with virtual keys, budgets, attribution, provider access controls, no-retention prompt policy, and reconciliation-ready usage records.

Explore AI Gateway
BYOK

BYOK for Model Provider Access

Centralize OpenAI, Anthropic, Gemini, Vertex AI, and Bedrock credentials behind encrypted controls while giving developers governed virtual-key access through Cloptima.

Explore BYOK
Budgets

LLM Budget Controls

Set model spend, token, and request limits by team, app, environment, user, key, provider, and agent workflow.

Explore Budgets
Analytics

Model Spend Analytics

Analyze AI spend by provider, model, team, app, environment, user, session, run, and custom business dimensions.

Explore Analytics
Reconciliation

LLM Provider Bill Reconciliation

Compare gateway and telemetry usage against provider billing exports with reconciliation-ready cost records.

Explore Reconciliation
Unit Economics

AI Unit Economics

Measure LLM cost per customer, workspace, agent run, ticket, document, transaction, and product workflow.

Explore Unit Economics
Agents

AI Agent Cost Controls

Track and govern agent sessions, retries, loops, tool calls, MCP workflows, and autonomous runs before they create runaway AI spend.

Explore Agents
Shadow AI

Shadow AI Discovery

Find unowned LLM usage, direct provider calls, and AI spend that is not mapped to teams, apps, or approved governance paths.

Explore Shadow AI
Private Gateway

Private LLM Edge Gateway

Run AI gateway enforcement closer to workloads with local or VPC deployment, locally encrypted credentials, and signed policy snapshots.

Explore Private Gateway
Guardrails

LLM Guardrails and Guardrail Cost Governance

Detect prompt injection, jailbreaks, PII, and secrets in LLM traffic, and govern the cost of running those checks with policy-based bypass, approval workflows, and audit trails.

Explore Guardrails
Exact Cache

Exact Response Cache for LLM Requests

Serve identical, policy-approved LLM requests from cache instead of the provider, with observe and enforce modes, namespace isolation, and a blast-radius preview before you turn enforcement on.

Explore Exact Cache
Semantic Search

Semantic Search and Semantic Response Cache

Search across LLM traces, prompts, tools, and policy violations by meaning, and serve equivalent — not just identical — requests from a policy-gated semantic cache with eval-scored safety checks.

Explore Semantic Search
Prompt & Eval Registry

Prompt Registry, Datasets, and Eval-Gated Releases

Version prompts, manage evaluation datasets, run deterministic and LLM-as-judge evals, and gate model or prompt changes behind a passing eval score with canary rollout and rollback.

Explore Prompt & Eval Registry
Contract Pricing

Contract Pricing and Enterprise Rate Overrides

Apply customer- or org-specific contract rates, commitments, and credits on top of list pricing, with a server-authoritative preview before approval and an immutable audit history behind every override.

Explore Contract Pricing

Compare AI FinOps Options

Solutions by Team

FAQ

Put LLM Spend Under the Same Discipline as Cloud Spend

Start with telemetry, governed gateway routing, or both. Keep model usage, ownership, and finance reporting in one operating model.

No credit card required
5-minute setup
Free trial