Open source · Self-hosted · v0.1.0

AI infrastructure costs,
under control.

Full observability for every AI API call — track cost, latency, and token usage across all your providers in real-time. Instrument with one line of code, gain instant visibility into your AI infrastructure, and scale to enterprise-grade routing, budget enforcement, and governance when you need it. Free, open source, and fully self-hosted.

Get Started Free See how it works Enterprise Demo

stacksense · live

Today's AI spend

$3.84

Total tokens today

1.24M

Avg latency (p95)

312ms

Providers active

Cost this month

$47.20

Budget remaining

82%

# one line to start

client = ss.monitor(client)

# that's it.

Two-tier architecture

Visibility first. Optimization when you need it.

Start with complete observability for free. Add enterprise-grade routing, budget enforcement, and governance as you scale.

Open Source

Visibility Layer

Complete observability, free and self-hosted. The foundation every AI team needs.

✓
Unified dashboard
Real-time cost, latency, and token metrics across all providers. Drill down by model, timeframe, or environment with interactive charts and tables.
✓
Zero-code instrumentation
Wrap any client in one line, no SDK changes required. Automatic capture of all metrics including errors, retries, and streaming responses.
✓
Multi-provider support
OpenAI (GPT-4, GPT-3.5), Anthropic (Claude), ElevenLabs (voice), Pinecone (vectors). Unified interface across all providers with automatic cost calculation.
✓
SQLite + PostgreSQL
Local dev with SQLite for zero-config testing. Production-ready PostgreSQL support with automatic schema migrations and connection pooling.
✓
Google OAuth + key vault
Encrypted API key storage with AES-256. Per-user accounts with Google OAuth, role-based access control, and audit logging. Fully self-hosted.
✓
Docker + Kubernetes
Production Docker Compose setup included. Kubernetes Helm charts in beta with horizontal scaling, health checks, and Prometheus metrics export.

Enterprise

Optimization Engine

Active cost control — route, enforce, govern. The CFO layer for your AI infrastructure.

✓
Dynamic model routing
Automatically route requests to the most cost-effective model based on prompt complexity, required quality tier, and latency SLAs. Save 40-60% on AI costs without sacrificing performance.
✓
Budget circuit breakers
Set per-team, per-feature, or global spending limits. Auto-downgrade to cheaper models or rate-limit when budgets are hit. Zero service disruption, full spend control.
✓
Token waste detection
ML-powered prompt scoring identifies inefficient patterns, redundant context, and bloated system prompts. Get specific recommendations to reduce token usage by 20-50%.
✓
Cross-vendor arbitrage
Real-time price monitoring across OpenAI, Anthropic, and others. Automatically shift traffic to maximize quality-per-dollar. Handle provider outages with instant failover.
✓
Governance + audit logs
Tamper-evident blockchain-style audit trail. Model allowlists, PII detection with automatic redaction, data residency enforcement. Full SOC 2 compliance support.
✓
Agent tracking
Track multi-step agentic workflows end-to-end. Total cost per run, infinite loop detection, per-task token budgets. Full step-by-step audit trail with tool usage breakdown.

The philosophy

Start with visibility — free, forever. Add optimization when your AI spend is worth controlling. Two tiers, one clean upgrade path.

Start for free Full comparison

Core capabilities

Everything you need out of the box.

StackSense gives you complete control over your AI infrastructure from day one. No partial solutions, no "coming soon" for critical features.

📊

Real-Time Cost Tracking

Automatic cost calculation per model and provider. Live updates every second with sub-cent precision. Historical trends, cost breakdowns by feature, and instant alerts when spending patterns change.

⚡

Latency & Performance Monitoring

Track p50, p95, p99 latencies across all requests. Automatic slow query detection, streaming vs batch performance analysis, and provider comparison dashboards to identify the fastest endpoints.

🔢

Token Usage Analytics

Granular input/output token breakdown. Track token efficiency over time, identify expensive patterns, and get automatic recommendations for prompt optimization and context reduction.

🔍

Error & Retry Analysis

Comprehensive error tracking with automatic categorization. Monitor rate limits, failed requests, and timeout patterns. Intelligent retry analysis helps identify wasteful retry loops costing you money.

📈

Multi-Timeframe Views

1H, 24H, 7D, and 30D views with intelligent data aggregation. Compare across periods, spot trends early, and understand seasonal usage patterns. Custom date ranges for detailed analysis.

🌍

Multi-Environment Support

Separate tracking for development, staging, and production. Environment-specific budgets, isolated metrics, and the ability to test cost impact before deploying to production.

🔐

Encrypted Key Management

AES-256 encrypted storage for all API keys. Per-user key vaults with role-based access control. Automatic key rotation reminders and audit logs of who accessed what, when.

📡

Live Monitoring Dashboard

WebSocket-powered real-time updates with zero polling overhead. Live alerts for budget thresholds, error spikes, and latency degradation. System health checks with Prometheus metrics export.

Built-in support

Plug into your entire stack.

🤖

OpenAI

Live

🧠

Anthropic

Live

🔊

ElevenLabs

Live

🌲

Pinecone

Live

☁️

AWS Bedrock

Building

🧬

Google AI

Building

☸️

Kubernetes

Beta

📊

Grafana

Planned

📈

Datadog

Planned

🐋

Docker

Live

🗄️

PostgreSQL

Live

🪶

SQLite

Live

Quick start

From zero to full visibility in 2 minutes.

Three simple steps to instrument your entire AI stack. No configuration files, no complex setup, just wrap and go.

STEP 1 · INSTALL

# Install StackSense
pip install stacksense

# Or with Docker
docker compose up -d

PyPI package or Docker Compose. Zero dependencies, works with Python 3.8+.

STEP 2 · WRAP CLIENT

# Wrap any AI client
import stacksense as ss

client = OpenAI()
client = ss.monitor(client)

# That's it!

One line wraps your client. All methods automatically tracked with zero code changes.

STEP 3 · OPEN DASHBOARD

# Start dashboard
make dashboard

# Open browser
http://localhost:5000

# See all metrics live!

Dashboard auto-starts on port 5000. Real-time metrics, zero configuration required.

Full example

from openai import OpenAI
import stacksense as ss

# Initialize and wrap your client
client = ss.monitor(OpenAI())

# Use normally - every call is automatically tracked
response = client.chat.completions.create(
    model="gpt-4",
    messages=[{"role": "user", "content": "Hello!"}]
)

# ✅ Cost, latency, tokens → tracked automatically
# ✅ Visible in dashboard at http://localhost:5000
# ✅ No configuration needed

Use cases

Built for teams shipping AI products.

From indie hackers to enterprise engineering teams, StackSense helps you understand and control AI costs from day one.

🚀

Startups & Indie Hackers

Track every dollar spent on AI from day one. Know exactly which features cost what. Catch runaway costs before they hit your runway. Free tier means zero overhead.

🏢

Enterprise Teams

Multi-team visibility with per-project budgets. Compliance-ready audit trails. Dynamic routing saves 40-60% on AI spend. SOC 2 ready governance engine included.

🤖

AI Agent Developers

Track entire agentic workflows end-to-end. Catch infinite loops before they drain your budget. Per-task token limits. Full audit trail of every tool call and decision.

📊

Data Science Teams

Compare model performance vs cost in real-time. A/B test prompts with instant ROI analysis. Track embeddings, vector searches, and inference costs in one place.

💼

FinOps & Platform Teams

Centralized AI cost management across all teams. Chargeback by department. Budget forecasting with trend analysis. Prometheus metrics for existing monitoring stack.

🎓

Researchers & Academics

Track research experiment costs. Compare LLM providers for your specific workload. Self-hosted means your data stays private. Open source means full control and customization.

Enterprise tier

From observability to active control.

Six capabilities that transform StackSense from a dashboard into the economic layer of your AI infrastructure.

⚡

Dynamic Model Routing

Route each prompt to the right model based on task complexity, cost thresholds, and latency requirements. Automatic fallback when quality permits.

Routing

🔬

Token Waste Detection

Score prompts for efficiency. Detect redundant context, bloated system prompts, and high retry rates. Get specific recommendations to cut spend.

Cost Intelligence

🛑

Budget Circuit Breakers

Per-team, per-feature, or global limits. When hit, StackSense auto-downgrades model tiers or rate-limits requests with no service disruption.

Budget Enforcement

⚖️

Cross-Vendor Arbitrage

Monitor real-time pricing and latency across providers. Shift traffic automatically to maximize value without degrading user experience.

Multi-Provider

📋

Governance & Audit

Tamper-evident logs of every AI call. Model allowlists, PII detection, data residency enforcement, compliance reporting for SOC 2.

Governance

🤖

Agent Tracking

Track agentic workflows end-to-end. Total cost per run, loop detection, task-level token budgets, full step-by-step audit trail.

Agent Intelligence

Full feature comparison

OSS vs Enterprise

Feature	🟢 Open Source	🔵 Enterprise
Cost & token tracking	✓	✓
Latency & error monitoring	✓	✓
Multi-provider dashboard	✓	✓
SQLite + PostgreSQL	✓	✓
Docker / self-hosted	✓	✓
Kubernetes integration	✓ Beta	✓ Full
Google OAuth + accounts	✓	✓ + SSO/SAML
Dynamic model routing	—	✓
Budget circuit breakers	—	✓
Token waste detection	—	✓
Cross-vendor arbitrage	—	✓
SLA-aware routing	—	✓
Agent workflow tracking	—	✓
Enterprise policy engine	—	✓
Audit logs & governance	—	✓
AI unit economics	—	✓
Support	Community	Enterprise SLA
License	Open Source	Proprietary
Price	Free	Custom

Get started

Choose your path.

Open Source

Start monitoring in 5 minutes.

pip install stacksense — wrap your clients, see your costs. No signup, no credit card, no limits. Fully open source and self-hosted.

Get on GitHub View Live Dashboard →

Enterprise

Ready to optimize at scale?

Talk to us about dynamic routing, budget enforcement, and governance. We'll show you exactly how much you're leaving on the table.

Book an Enterprise Demo Compare all features →

AI infrastructure costs,under control.

Visibility first. Optimization when you need it.

Everything you need out of the box.

Plug into your entire stack.

From zero to full visibility in 2 minutes.

Built for teams shipping AI products.

From observability to active control.

OSS vs Enterprise

Choose your path.

AI infrastructure costs,
under control.