Now tracking 163 services across 7 LLM providers

Know your true cost.
Per customer. Per agent. Per dollar.

Agent Unit Economics, made measurable. Track every dollar your AI agents spend — the vector DB queries, retries, and tool calls that LLM observability misses. One line of code, attributed to every customer and every task.

Free up to 5,000 tasks/month · No credit card · SDKs MIT-licensed

dexcost dashboard — cost overview with per-customer and per-task breakdown

Cost calculator

How much are your agents really costing?

Configure

10010,000

Monthly Cost Breakdown

LLM Costs$180.00/mo
OpenAI GPT-4o$112.50
Anthropic Claude$67.50

← You probably track this

Tool Costs$972.45/mo
Pinecone$720.00
Tavily Search$240.00
Firecrawl$12.45

← Do you track these?

Total monthly cost$1,152

500 tasks/day × 30 days

~$57.62/mo in retry cost is potentially recoverable

Start tracking for free

How it works

Three steps to cost clarity

Drop in the SDK, let dexcost capture every event, see the full picture in your dashboard.

1

One line of code

import dexcost dexcost.init(api_key="...")

Auto-instruments LLM calls and HTTP requests. Works with any framework.

2

163 services tracked

From Tavily to Pinecone. Costs extracted from API responses automatically. No manual tracking.

OpenAIAnthropicGeminiTavilyPineconeE2B+53 more
3

See everything

Per customer. Per agent. Per task. Retry costs. Anomalies. Profitability. All in one dashboard.

Customer A$12.40
Customer B$84.20
Customer C$3.10

Capabilities

Built for Agent Unit Economics

Full-stack cost capture

LLM tokens, API tool calls, compute seconds, vector DB queries — every cost in one event stream, automatically.

Per-customer attribution

Know exactly how much each customer costs. Spot unprofitable accounts before they drain margin.

One line to start

Python, TypeScript, Go, and Rust SDKs. Auto-instruments 7 LLM providers. No code changes to your agent logic.

Retry cost recovery

Retries are tracked automatically. See how much you spend on failures, timeouts, and rate limits — and recover it.

Spend caps where attribution lives

Set per-customer, per-project, or per-task-type budgets and get alerted before they're breached — not after. Most observability tools stop at the workflow level. We attribute alerts to the customer eroding margin.

Confidence scoring

Every cost event tracks its confidence level — exact, computed, estimated, or unknown. No black boxes.

7 LLM providers · 163 services · Zero config

OpenAIAnthropicGoogle GeminiAWS BedrockCohereLiteLLMVercel AI

Open source · MIT licensed · Backed by Dexwox Innovations

7 LLM providers163 services4 SDK languages

Stop guessing. Start tracking.

See exactly where every dollar goes — per customer, per agent, per task. Set up in 5 minutes.

Free up to 5,000 tasks/month · No credit card required