Dashboard — Costiva

You're viewing a live demo with sample data — this is the real Costiva dashboard. Create a free account to connect your own Anthropic key and see your real costs.

Total spend (60d)

$1,132.86

▼ 25.4% vs prev 60d

Total tokens (60d)

21,397,791

$52.94 / 1M tokens

Cache hit: 58% · ~$29.36 saved

Projected MTD spend

$380.46

linear from month-to-date

Active surface

3 models · 2 workspaces

in the selected range

AI ROI by KPI 4 KPIs · 60d

Net gain / loss

+$1,667.43

across 3 priced KPIs

Value gained

$2,850.00

Spend mapped to KPIs

$1,182.57

of $1,132.86 total

Health

3 winning · 0 leaking

Spend not paying off

Follow-ups per rep / week

Watch

$292.87

Top performers

Median first-response time	+$927.60
Ticket deflection rate	+$543.80
Time to first follow-up	+$196.03

A unit/KPI mapped to several KPIs counts under each, so these totals can overlap. Net needs value tracking ($/unit) — set it on a KPI's Track value & ROI page.

Tools Free tools included on every plan

Plan

Instant · no queue

Workload Cost Estimator

Estimate your monthly Claude API spend before you build — with prompt caching and Batch API discounts.

Open

Instant · no queue

Model Picker

Which AI model should you use? Answer a few questions and get a transparent, explainable recommendation across Claude, GPT, and Gemini.

Open

Instant · no queue

Token Counter

Count real tokens for your text, per Claude model — different models tokenize differently.

Open

Build

AI-generated

CLAUDE.md Generator

Generate a tailored CLAUDE.md for your project in under a minute.

Open

Instant · no queue

MCP Config Wizard

Build a valid claude_desktop_config.json or .mcp.json in a few clicks — no hand-written JSON.

Open

AI-generated

Skill File Generator

Generate a SKILL.md with a description that triggers at the right moments — and never at the wrong ones.

Open

Test

AI-generated

CLAUDE.md Auditor

Score your CLAUDE.md on actionability, hard rules, signal density, completeness, and agent ergonomics.

Open

AI-generated

Eval Set Generator

Turn a feature description into a starter eval set: concrete test cases, a grading rubric, and a pass bar.

Open

AI-generated

Prompt Auditor

Score your system prompt on structure, specificity, edge cases, token efficiency, and cacheability.

Open

AI-generated

Skill Auditor

Score your SKILL.md on trigger precision, instruction completeness, output contract, and scope discipline.

Open

Govern

AI-generated

AI Usage Policy Generator

Generate a practical internal AI usage policy your employees will actually follow.

Open

Hypotheses View all

Hypothesis	KPIs	Created
Support deflection with an AI assistant We believe that drafting first replies to support tickets with an AI assistant for our support team will reduce average first-response time this quarter.	2	Jul 8, 2026	Manage KPIs
Faster sales follow-ups with Claude We believe that drafting follow-up emails with Claude for the sales team will speed up time to first follow-up within 8 weeks.	2	Jul 25, 2026	Manage KPIs

Recent assets View library

CLAUDE.md — Acme Support Bot v2 · Sep 2
Ticket Triage Skill v1 · Aug 29

Cost over time (USD)

Token usage over time (stacked)

Cost by model

Cost by workspace

Cost by type

Type	USD
`uncached_input_tokens`	$475.80
`output_tokens`	$521.12
`cache_read_input_tokens`	$79.30
`cache_creation_input_tokens`	$56.64

Usage by service tier tokens — priority cost is not in the cost endpoint

Tier	Tokens
standard	17,546,188
batch	3,851,602

Usage by context window

Context window	Tokens
200k	15,192,431
1M	6,205,359

Data residency tokens by inference geo

Geo	Tokens
not_available (pre-Feb-2026 models)	11,768,785
us	9,629,005

Fast mode vs standard tokens by speed

Speed	Tokens
standard	19,258,011
fast	2,139,779

Rate limit capacity estimated peak load over last 24h vs configured per-minute limits

Group	Models	Input/min	Peak in	Input %	Output/min	Peak out	Output %	Req/min
model_group	3 models	2,000,000	1,362,000	68.1%	400,000	208,000	52.0%	4,000

Utilization = estimated peak per-minute token load vs the configured limit. Request-per-minute limits show no utilization because the usage report measures tokens, not request counts. Highlighted rows are at/above 80%.

Top API keys

API key	Hint	Workspace	Tokens
Production key	`sk-ant-…a1b2`	Production	12,410,718
Support bot key	`sk-ant-…9f7c`	Support Bot	6,633,315
Batch jobs key	`sk-ant-…44de`	Production	2,353,757