One AI endpoint. Clear control of spend.
Connect your own provider keys, manage application access with Clara virtual keys, and understand AI usage and spend through one OpenAI-compatible endpoint.
- Your own provider keys
- Scoped virtual keys
- Usage and spend visibility
This month
- Spend · USD
- 1,248
- Requests
- 18,420
Providers
- OpenAIConnected
- AnthropicConnected
Virtual key
Support app
Sample usage
- Support
- Provider: OpenAI
- Spend · USD: 642
- Documents
- Provider: Anthropic
- Spend · USD: 384
- Operations
- Provider: OpenAI
- Spend · USD: 222
Entity · country · currency
Planned reportingReporting by legal entity, country and functional currency is planned.
Synthetic data for illustration. No customer activity is shown.
Explore the AI model ecosystem
- OpenAI
- Anthropic
- Meta
- Mistral AI
- DeepSeek
- xAI
- Cohere
- Qwen
Provider names illustrate the model ecosystem. Enabled providers and models are confirmed during beta onboarding.
The problem
Buying AI from Latin America shouldn't be this hard
AI providers were built for companies that pay in dollars from the United States. Everyone else pays a hidden tax in fees, paperwork and downtime.
Dollar invoices, local budgets
Providers bill in USD. FX spreads and taxes on international payments, such as IOF, inflate every invoice, and no forecast survives a currency swing.
Declined cards, stalled products
Banks flag recurring charges from abroad. When the card is declined, API access stops, and your product stops with it.
No valid tax invoice
A foreign receipt is not a CFDI or a nota fiscal. Finance chases paperwork every month and deductibility is at risk.
Keys and costs everywhere
Every team opens its own account with every provider. Nobody can say how much the company spends on AI, or on what.
Frontier prices for simple tasks
Most requests don't need the most expensive model. Without routing, classification, extraction and short answers all go to it anyway.
Benchmarks that only speak English
Public leaderboards rarely measure Spanish or Portuguese. Your customers notice the difference, even when the benchmark doesn't.
How it works
From provider keys to visible usage
The product workflow connects your existing AI accounts to controlled access and usage reporting. Contact us about beta access.
- 01
Connect your providers
Bring your own provider keys and keep your existing provider accounts.
- 02
Create Clara virtual keys
Give each application its own key and choose the models it can access.
- 03
Understand usage and spend
Send supported requests through one endpoint and review model usage and costs in USD.
Optimization roadmap
Lower costs, with guardrails on quality
Explore model routing, context efficiency and the controls that can make AI spend easier to manage. These capabilities are part of the product direction.
Explore the optimization and finance capabilities we are developing. Availability is confirmed for each beta setup.
Spend less
Smart routing
Explore choosing models by task complexity, language, latency target and cost ceiling.
Context optimization
Planned context controls can reduce redundant input and take advantage of provider-side prompt caching.
Response caching
Explore reusing eligible responses so repeated work does not require a fresh model call every time.
Language on/off-ramps
An experimental path for eligible Spanish and Portuguese workloads: compact input, then restore the original language.
Stay in control
Quality guardrails
Compare candidates against a task-specific quality bar and keep critical flows pinned to an approved model.
Automatic fallbacks
Provider-aware retries and fallback policies are part of the reliability direction; the enabled policy depends on the setup.
Budgets and hard limits
Budget controls by key, team or project are planned, with thresholds and hard limits confirmed for each pilot.
Cost analysis
Start with model, token and USD cost visibility; richer route explanations and savings analysis are planned.
Inside a routing decision
See the trade-off, request by request
An animated example of model choice, cost and latency. Quality scores and latency are synthetic; this illustrates the optimization direction.
Request
model: "clara/auto"
prompt: "Summarize this supplier contract in 5 bullet points"
Candidates
- Claude Opus 5.5
- Quality
- 98
- Cost
- $0.0208
- Latency
- 6.9 s
- Gemini 3.8 FlashSelected
- Quality
- 95
- Cost
- $0.0078
- Latency
- 2.6 s
- GPT-6 LunaBelow quality bar
- Quality
- 90
- Cost
- $0.00052
- Latency
- 1.9 s
This request
$0.0078
vs. $0.0208 on Claude Opus 5.5
Payment benefit scenario
+5%
Savings that stack
Model the savings before you switch.
Explore a scenario with up to 40% optimization and a 5% payment benefit. Both are assumptions in this calculator, not measured results or a guaranteed offer.
Optimization comes first
The payment scenario applies after optimization: 40% and then 5% combine to 43% in the worked example.
Change the assumptions
Adjust your monthly spend and expected optimization to see the effect on monthly and yearly totals.
Connect savings to spend visibility
Use the estimate to start a conversation about workloads, ownership and the product capabilities available to your team.
Estimate your savings
Modeled monthly savings: $4,800/mo. Modeled yearly savings: $57,600/yr.
- Modeled optimization savings
- $4,000/mo
- Payment benefit scenario (5%)
- $800/mo
- Modeled monthly savings
- $4,800/mo
Modeled yearly savings
$57,600/yr
Illustrative assumptions, not a guarantee. The 5% payment scenario applies after optimization; actual terms are confirmed during onboarding.
For finance
See the spend behind your AI usage
Start with model usage and costs in USD. Bring engineering and finance into the same conversation about which models are used and what they cost.
- Visibility into providers and models
- Request and token usage alongside costs
- Spend displayed in USD
Next: reporting for your organization
Reporting by legal entity, country and functional currency is planned. The preview illustrates that direction; it does not represent local billing or tax invoicing available today.
Built for Latin America
Made here, for companies that operate here
AI spend crosses borders; budgets, accounting and language remain local. The roadmap brings those needs into the same product.
Explore the optimization and finance capabilities we are developing. Availability is confirmed for each beta setup.
| Country | Currency | Local document |
|---|---|---|
| Mexico | MXN | CFDI 4.0 |
| Brazil | BRL | Nota fiscal (NFS-e) |
| Colombia | COP | Factura electrónica DIAN |
| Rest of the region | USD | Commercial invoice |
Local-currency billing
Planned: report and manage AI spend in MXN, BRL or COP alongside the original USD cost. Local billing remains a future capability.
Tax-valid invoices
Planned: workflows for CFDI in Mexico, nota fiscal in Brazil and factura electrónica in Colombia, subject to country-specific availability.
Spanish- and Portuguese-aware routing
The evaluation direction includes Spanish, Portuguese and regional variants, so model choice can reflect the language your customers use.
Lower latency in the region
Regional latency is a consideration when evaluating provider endpoints. Residency and local capacity depend on the selected provider and setup.
Privacy under local law
Data-processing requirements are reviewed for each setup, including the needs of companies operating in Brazil, Mexico and Colombia.
Support in your language
Discuss onboarding and support in Spanish, Portuguese or English with the Clara team.
One platform, two teams
Engineering ships faster. Finance finally sees the bill.
Developers
Keep the SDK pattern your team knows
Configure your gateway URL, Clara key and an enabled model. Move one workload first and check its usage.
Integration pattern: use the gateway URL and enabled model provided during onboarding. The public landing page does not serve an inference API.
- An OpenAI-compatible pattern for supported models and request features
- Keep gateway configuration in environment variables, outside application code
- Use the enabled model and integration guidance provided for your beta setup
# Same SDK pattern, with your onboarding configuration
import os
from openai import OpenAI
client = OpenAI(
base_url=os.environ["CLARA_ROUTER_BASE_URL"],
api_key=os.environ["CLARA_ROUTER_API_KEY"],
)
response = client.chat.completions.create(
model=os.environ["CLARA_ROUTER_MODEL"], # enabled model from your setup
messages=[{"role": "user", "content": "Classify this support ticket by urgency"}],
)
Comparison
Why not just go direct?
| Feature | Clara Router | Direct provider accounts | Global routers |
|---|---|---|---|
| Clara payment discount | Confirm terms | Provider terms | Router terms |
| Billing in MXN, BRL and COP | Planned | Provider billing | Router billing |
| Local tax invoices | Planned | Provider documents | Router documents |
| One API for all models | Supported models | Per-provider endpoint | Check model coverage |
| Cost-optimized routing and fallbacks | Planned | Application integration | Check routing policies |
| Routing informed by Spanish and Portuguese evals | Planned | Evaluate each model | Check language coverage |
| Budgets by cost center, with approvals | Planned | Account-level controls | Check control scope |
| Accounting and ERP sync | Planned | Accounting integration | Accounting integration |
| Support in Spanish and Portuguese | Confirm terms | Check support languages | Check support languages |
A guide to the questions worth comparing. Planned Clara capabilities are marked; provider terms and features vary by account, country and product.
Trust and control
Keep access deliberate and reporting clear
Know which credentials power a request and what the reporting represents.
Metadata for gateway reporting
Gateway usage reporting records request metadata, such as model, token usage and cost, rather than prompt or response text.
Your provider accounts
Bring your own provider keys. Your company manages the provider accounts and the credentials that connect them.
A demo with synthetic data
The public demo uses synthetic data. It does not collect provider keys or send requests to AI providers.
Pricing
Find the right beta setup for your team
Start with your provider accounts and a clear view of model costs. Discuss payment benefits, volume and support before moving traffic.
Beta commercial preview. Model costs, payment benefits, eligibility and service terms are confirmed with your team.
Clara customers
Explore the beta
Discuss how Clara Router can fit the accounts and workloads your company already manages.
- Bring your own provider keys
- Clara virtual keys and supported models
- Usage and USD cost visibility
- Discuss future finance workflows
- Payment benefits subject to confirmed terms
High volume
Discuss your workload
For teams evaluating committed usage, capacity needs and closer support.
- Review models and expected traffic
- Discuss volume and commercial terms
- Confirm rate limits and capacity
- Review data-processing requirements
- Plan a gradual workload rollout
New to Clara
Explore the ecosystem
Learn about Clara and discuss the access requirements for your team.
- Corporate cards and payments
- Company-wide spend management
- Router beta access confirmed separately
FAQ
Before you get started
Can I use Clara Router today?
Clara Router is in beta. Explore the public demo, then contact us to discuss access and the scope available to your team.
Do we bring our own provider keys?
Yes. The product starts with your own provider accounts and keys. Clara virtual keys then give your applications scoped access through the router.
Is the endpoint OpenAI-compatible?
Yes, the product uses an OpenAI-compatible endpoint for supported models. Check the product integration guide for the models and request features enabled for your setup.
How is spend reported?
Current usage reporting shows model usage and costs in USD. Reporting by legal entity, country and functional currency is planned; the landing page shows a clearly marked preview.
Does the demo connect to the live product?
No. The public demo runs with synthetic examples and does not use your provider credentials or call AI providers. Simulator results illustrate assumptions rather than actual customer usage.
Bring clarity to your AI access and spend
Explore the demo, then talk with us about connecting your providers and giving your team controlled access.