Skip to content

Dashboard Guide

The AgentCost dashboard provides seven intelligence views for understanding and controlling your AI costs.

Launching the Dashboard

agentcost dashboard
# → http://localhost:8500

Or with custom port:

AGENTCOST_PORT=8100 agentcost dashboard

Seeding Demo Data

To explore the dashboard immediately:

curl -X POST http://localhost:8500/api/seed \
  -H "Content-Type: application/json" \
  -d '{"days": 14, "clear": true}'

Or use the script:

python scripts/seed_sample_data.py --days 14 --clear

Views

Overview

The landing page shows key metrics at a glance: total spend, call volume, error rate, and cost-over-time charts. Use the project filter to drill into specific applications.

Cost Breakdown

Detailed spend analysis by model, project, and provider. Shows which models consume the most budget and identifies cost trends over time.

Forecasting

Predicts future costs using three methods:

  • Linear — Simple trend extrapolation
  • EMA — Exponential moving average (adapts to recent changes faster)
  • Ensemble — Weighted combination (recommended)

Includes budget exhaustion prediction: "At current rates, your $1,000 monthly budget will be exhausted by March 15th."

Optimizer

Analyzes your usage patterns and recommends model substitutions:

  • Which calls use expensive models for simple tasks?
  • What savings would you get from switching gpt-4o → gpt-4o-mini for classification?
  • Overall efficiency score for each project

Analytics

Deep-dive analysis:

  • Token efficiency — Cost per 1K tokens by model
  • Top spenders — Highest-cost agents, projects, or models
  • Error analysis — Error rates by model and type
  • Chargeback reports — Cost allocation by team/project

Estimator

Pre-call cost estimation. Enter a prompt and compare estimated costs across 2,610+ models before making expensive API calls.

Models Explorer

Search and filter the full model pricing database:

  • Real-time search by model name across 2,610+ models
  • Filter by provider — OpenAI, Anthropic, Google, Groq, DeepSeek, Bedrock, Azure, and 30+ more
  • Filter by tier — Economy, Standard, Premium, Free
  • Filter by cost range — Set max input cost per 1M tokens
  • Filter by context window — Minimum context size
  • Click any model for detailed pricing, tier classification, and capability info (vision, tools)

The Models Explorer is powered by the vendored pricing API at /api/models, which serves data from the same 2,610-model database used for cost calculation.

API Endpoints

All dashboard views are powered by REST endpoints you can use directly:

Endpoint Description
GET /api/summary Overview metrics
GET /api/cost/by-model Cost breakdown by model
GET /api/cost/over-time Time-series cost data
GET /api/forecast/{project} Cost predictions
GET /api/optimizer/{project} Optimization recommendations
GET /api/analytics/{project}/summary Usage analytics
POST /api/estimate Cost estimation

See the API Reference for full documentation.