Now tracking OpenAI, Anthropic & more

The FinOps
platform for Agents.

The FinOps layer for AI agents. Track spending, enforce budgets, and approve costs in real-time. Two config changes. No SDK rewrite.

Start Free Read the Docs

$2M+AI spend tracked

50msavg latency

99.9%uptime

Live

Total Spend (24h)

+12.4% from yesterday

API Requests

847K total

Avg Cost/Request

$0.015

-8.2% optimized

Spend by Model

Last 7 days

MonTueWedThuFriSatSun

Live Transactions

Streaming...

gpt-5.4

2,847 tokens$0.08542ms ago

claude-opus-4.6

1,293 tokens$0.038715ms ago

gpt-5.4-mini

8,472 tokens$0.012748ms ago

claude-sonnet-4.6

3,918 tokens$0.0235102ms ago

gpt-5.4

1,847 tokens$0.0554156ms ago

Real-time tracking

Per-request cost tracking.Down to the token.

Every request logged with input, output, cached, and reasoning tokens. Costs calculated automatically using the latest provider pricing. No estimation, no surprises.

Total Spend

$0.0000

Input

Output

Cached

Monthly BudgetACTIVE

$0/ $100

Hard limits

Budget enforcement.Before the request leaves.

Set hard spending ceilings per key, team, or project. The proxy returns 429 before the request ever reaches OpenAI or Anthropic. No overages, no surprises, no arguments with finance.

Auto-protection

Detect runaway loops.Auto-circuit-breaker.

Set velocity thresholds for $/minute spend rates. When an agent goes haywire, NullSpend automatically trips the circuit breaker before your bill does.

SPEND VELOCITY

LIVE

$/minMonitoring

X-NullSpend-Tags:

productionstagingteam:mlfeature:search

production

$42.18

staging

$28.54

team:ml

$15.92

feature:search

$63.41

Attribution

Tag everything.Attribute costs anywhere.

Add a single header to attribute costs to teams, environments, or features. Set default tags on API keys. Finally know where your AI spend is going.

MCP Native

Track every tool call.Cost per action, not just tokens.

Your agents call tools across MCP servers. We track each call, attribute costs, and give you visibility into what your agents are actually doing.

MCP Tool Calls

$0.000

Estimated Cost$24.50

Modelgpt-5.4

Tokens~128,000

Awaiting Approval

Agent paused. Human decision required.

Human control

Agents wait.Humans decide.

Approval workflows for high-cost or sensitive operations. Set cost thresholds, require sign-off for certain models, or flag specific prompt patterns. The agent pauses until a human approves.

Up and running in two minutes

Point your SDK

Change your base URL to proxy.nullspend.dev. One environment variable.

Add one header

X-NullSpend-Key authenticates requests. Your provider keys pass through untouched.

See costs instantly

Every request is tracked, budgeted, and visible in your dashboard and via webhooks.

Three ways to connect. Zero SDK rewrites.

Use the proxy, SDK, or Claude Agent adapter — all feed the same dashboard.

import OpenAI from "openai";

const client = new OpenAI({
  baseURL: "https://proxy.nullspend.dev/v1",
  defaultHeaders: {
    "X-NullSpend-Key": process.env.NULLSPEND_API_KEY,
  },
});

Simple, transparent pricing

Pay for what you use. Upgrade as your AI spend grows.

Free

$0/mo

Full cost tracking and budget enforcement for individuals.

Up to $5K/mo proxied spend
3 budgets with full enforcement
Velocity & session limits
30 day data retention
10 API keys
2 webhook endpoints

Get Started

Popular

Pro

$49/mo

Unlimited budgets and deeper analytics for production.

Unlimited team members
Up to $50K/mo proxied spend
Unlimited budgets
90 day data retention
Unlimited API keys
25 webhook endpoints
Priority support