How do I track costs per customer or user?

Pass a `cost_tracking_id` in the `extra_body` parameter when making API calls. Costbase captures this ID and provides aggregated analytics on your dashboard. Use cases: • SaaS billing per customer • Per-user cost attribution • Project-based billing • Team cost allocation • Feature-level cost analysis Cost tracking is available on all plans.

What are Guardrails and why do I need them?

Guardrails automatically detect and mask sensitive data in real-time—far beyond basic PII. We scan for **31 pattern types** across **100 regex variants** in <5ms. **What we detect:** • PII (SSN, credit cards, emails, phones) • PHI/HIPAA (medical records, health insurance IDs) • Cloud credentials (AWS, GCP, Azure) • LLM API keys (OpenAI, Anthropic, Cohere, etc.) • Database connection strings • Cryptographic keys (RSA, SSH, PGP) • Developer tokens (GitHub, GitLab, Slack) • Payment data (Stripe, crypto addresses) **Critical for:** • Healthcare (HIPAA compliance) • Finance (PCI-DSS, credit card protection) • SaaS/DevOps (prevent API key leaks to LLMs) • Legal (client confidentiality) • Enterprise (comprehensive DLP for AI) Built directly into the gateway—no additional tooling required. Available on Free+ tiers.

Up to 97% on LLM costs. Our real test results show 96.8% cost reduction through intelligent routing and semantic caching combined. Individual features: semantic caching saves 15-30%, intelligent routing saves 10-50%. Your actual savings depend on your usage patterns.

How hard is integration?

Drop-in replacement for OpenAI SDK. Just change your base URL and you're done.

What providers do you support?

OpenAI, Anthropic, Google, Cohere, Groq, Mistral, DeepSeek, xAI. 114+ models total across chat, embeddings, image generation, and audio transcription.

Do you store my API keys?

For Managed Cloud plans, yes - we store them securely using AES-256 encryption to route your requests. We never share them. For Self-Hosted Enterprise, you manage your own keys in your own infrastructure and we never see them.

Do you store my prompts and responses?

We store truncated previews (~50-100 characters) of prompts and responses for observability, debugging, and cost tracking. **With Guardrails (Free+):** Sensitive data is automatically masked before storage. SSNs become [SSN], API keys become [API_KEY], etc. Only masked previews are stored. **Without Guardrails:** Original text is truncated and stored. Consider upgrading to Free+ for automatic masking. **Retention:** • Free: 3 days • Pro: 30 days • Team: 90 days • Enterprise: 365 days Your provider API keys are encrypted using AES-256-GCM. We never train models on your data.

How does semantic caching work?

We use embedding-based similarity to detect similar prompts (85%+ similarity) and return cached responses instantly. Zero API calls, zero cost, ~5ms latency. Unlike simple string matching, semantic caching understands meaning. "What is AI?" and "Explain artificial intelligence" hit the same cache.

What's intelligent routing?

Set model: 'auto' and we pick the best model based on query complexity, cost targets, and performance needs. Three strategies: 1. Cost-based: Picks cheapest model 2. Semantic: Analyzes prompt complexity and routes accordingly 3. Performance-based: Routes to fastest models when latency matters Default is semantic (best balance of cost and quality).

How do budget limits work?

Set monthly request caps with hard stops. Real-time alerts at 50%, 80%, 90%, and 100% thresholds. Requests are blocked when you hit your plan limit until the next billing period. You're never surprised by unexpected bills.

Yes. Enterprise plan includes production-ready Docker, Kubernetes, and Terraform templates for AWS, GCP, Azure. Run Costbase in your own infrastructure with complete data control.

What if a provider goes down?

Automatic failover tries up to 3 providers before failing. Circuit breaker pattern prevents repeated failures. Your app stays up even when OpenAI goes down.

AES-256 encryption for all data at rest and in transit. Self-hosted option gives you complete control.

How is this different from Portkey or Helicone?

**Costbase optimizes costs. Others observe them.** | Feature | Costbase | Portkey | Helicone | |---------|----------|---------|----------| | Intelligent Routing (auto model selection) | ✓ | ✗ | ✗ | | Semantic Caching (built-in) | ✓ | Add-on | ✗ | | Guardrails (31 patterns, 100 variants) | ✓ | Basic | ✗ | | Zero Markup (BYOK) | ✓ | % fee | % fee | | Cost per Request in SDK | ✓ | ✗ | ✗ | **Key differences:** • **Intelligent routing** automatically selects the optimal model—GPT-4o-mini for simple queries, Claude for complex reasoning. Others make you pick manually. • **Semantic caching** is built-in, not an expensive add-on. ~5ms responses, $0 cost. • **Guardrails** detect 31 pattern types including cloud credentials, API keys, database strings—not just basic PII. • **Zero markup**: You pay providers directly via BYOK. We charge a flat subscription, not a % of your spend. Costbase makes you money. Others just show you where you spent it.

Cost Intelligence at a Price That Works for You

Intelligent routing, semantic caching, and PII/PHI guardrails that save money and protect sensitive data.

14-day free trial. No credit card required. OpenAI SDK compatible. Zero markup.

Start 14-Day Free Trial Calculate Your Savings

Three Plans. No Hidden Fees. Cancel Anytime.

Free

$0/month

Perfect for evaluation and side projects.

5K requests/month

Free forever*

All 8 providers (OpenAI, Anthropic, Google, Mistral, Cohere, xAI, Groq, DeepSeek)
114+ models across chat, embeddings, image gen, audio
Semantic caching (15-30% instant savings)
Intelligent routing (auto model selection)
PII/PHI Guardrails (31 pattern types, 100 regex variants)
Cost dashboard (view only)
BYOK (zero markup)
1 team member
1 project
3-day log retention

Start Free Trial

Best for:

Evaluation, side projects, prototypes

Pro

$39/month

Full cost control for indie devs and early-stage startups.

100K requests/month

14-day free trial

Everything in Free, plus:
Cost tracking by ID (SaaS cost attribution)
Budget limits & webhook alerts
Cost anomaly detection
Full cost analytics & trends
Rate limit alerts
Balance low alerts
3 webhooks (Slack/Discord/custom)
API key management (full + expiration)
MFA (TOTP-based 2FA)
RBAC (Admin, Developer, Viewer roles)
Model performance benchmarking
Automatic failover (1 retry)
Load balancing
99.5% uptime SLA
5 team members
3 projects
30-day log retention
Priority email support

Start Free Trial

Best for:

Indie devs and early-stage startups shipping AI features

Savings example:

Customer spending $5K/month on OpenAI saves $750-1,500/month through intelligent routing. Pro tier ROI: 15-30x.

Team

$299/month

Advanced controls for growth-stage teams.

1M requests/month

14-day free trial

Everything in Pro, plus:
Automatic failover (up to 3 retries)
Per-project budgets
Multi-threshold alerts
Cost forecasting
SLO violation alerts
10 webhooks
Alert cooldown control
Alert muting
Custom routing rules
Audit logging
99.9% uptime SLA
25 team members
10 projects
90-day log retention
Dedicated Slack/Discord support

Start Free Trial

Best for:

Teams with $20K-100K/month LLM spend

Savings example:

Customer spending $50K/month saves $7.5K-15K/month through intelligent routing and caching.

Enterprise

Custom

High-volume production workloads with enterprise requirements.

Unlimited requests/month

Everything in Team, plus:
Unlimited team members
Multiple organizations
Unlimited projects
SSO/SAML (Google, Okta, Azure AD, Auth0)
On-prem deployment
VPC deployment (AWS, GCP, Azure)
Multi-level budgets
Custom dashboards
Data export to data lakes
Custom log retention
Docker + Kubernetes + Terraform
99.95% custom SLA
24/7 priority support
Dedicated account manager
Custom development

Contact Sales

Best for:

Enterprises spending $200K+/month on LLMs

Savings example:

Typical savings: $30K-100K+/month through intelligent routing and caching at scale.

Compare All Features

See exactly what's included in each plan

Feature	Free	Pro	Team	Enterprise
Core Features
Monthly Requests	5K	100K	1M	Unlimited
Providers Supported	8	8	8	8
Semantic Caching
Intelligent Routing
Security
PII/PHI Guardrails
MFA (TOTP)
RBAC
Audit Logging
SSO/SAML
Analytics & Monitoring
Cost Tracking by ID
Budget Limits & Alerts
SLO Monitoring
Webhooks	0	3	10	Unlimited
Cost Anomaly Detection
Advanced Features
Automatic Failover		1 retry	3 retries	Custom
Support Level	Email	Priority Email	Dedicated Slack	24/7 + Account Manager
Cost Forecasting
Model Performance Benchmarking
Custom Routing Rules
Per-Project Budgets
Organization
Team Members	1	5	25	Unlimited
Projects	1	3	10	Unlimited
Log Retention	3 days	30 days	90 days	365 days
Deployment
Self-Hosted
VPC Deployment

How Costbase Saves You Money

1. Semantic Caching

15-30% instant savings

Same or similar prompts return cached responses in ~5ms with zero LLM cost.

2. Intelligent Routing

10-50% additional savings

Auto-selects the cheapest model that meets your quality needs.

3. Zero Markup

BYOK = No markup fees

Pay providers directly. We add intelligence, not cost.

4. PII/PHI Guardrails

Team+ tier exclusive

Automatically detect and mask PII/PHI before it reaches AI models.

Calculate Your Savings

Your monthly LLM spend: $5,000

$1K$200K

Estimated savings:

Semantic caching (20% avg)$1,000

Intelligent routing (30% avg)$1,500

No markup vs competitors (5%)$250

Total monthly savings:$2,750

Your Costbase investment:

Costbase cost:$49/month

Net savings:$2,701/mo

ROI:55.1x

Start Saving Today

Frequently Asked Questions

Ready to optimize your LLM costs?

Join hundreds of engineering teams using Costbase to cut costs and improve reliability.

Start Free Trial Contact Sales

PII/PHI Guardrails

AES-256 Encryption

99.97% Uptime

Zero Vendor Lock-in

OpenAI SDK Compatible

On-Premise Deployment