Savings from supply-chain efficiency
Volume purchasing, partner rates, reserved resources, and a lean platform margin reduce unit inference costs without model downgrades.
One Balance. One API Key. Every Connected Model.
CostRouter replaces separate provider subscriptions with cost-aware routing and usage-based billing. Fund one balance, access every connected model through one API Key, and pay only for actual usage.
New accounts receive $0.10 in introductory credits. Test routing, API compatibility, and billing first, then move to production when you are ready.
- One top-up
- One API Key
- Every connected model
- Pay per use
How to use
Connect your app or coding agent
01
AI app API integration
Use one API for GPT, Claude, Gemini, DeepSeek, and more with transparent usage-based pricing and cost-aware routing.
Apps
Backends
Workflows
One Base URL, existing SDKs, routed models
View guide
02
Start with coding agents
Route Codex, Claude Code, Gemini CLI, Cursor, and custom coding agents through your CostRouter API key.
CodexClaude CodeGemini CLICursor
View guide
Group options
Lower cost. Same requested model.
CostRouter lowers inference costs through volume purchasing and multi-channel sourcing while always requesting the model you specify. If one route is disrupted and another route to that model remains available, traffic can fail over. Overall availability still depends on the model provider and available partner channels.
Route-level failover
If one route is disrupted and another route to the requested model remains available, requests can fail over. This cannot prevent a provider-wide outage.
Latency-aware routing
Among compatible routes that meet reliability requirements, CostRouter considers live latency alongside price when selecting a path.
Incident service credits
Eligible CostRouter availability incidents are verified and compensated with account service credits under the applicable policy.
Request-level usage and cost visibility
Each request records the model, token usage, billing calculation, and final charge, making usage and spend easy to audit from the dashboard.
Standard groups
Delivered through partner channels such as AWS, Microsoft Azure, and Google Cloud, with more predictable throughput and sustained availability for production and business-critical workloads.
Value groups
Combines preferential rates from multiple compute and API partners, with each route verified for compatibility, model authenticity, and operational reliability. Ideal for individuals and small-to-medium usage.
Enterprise-grade reliability
For high-volume and business-critical workloads, we tailor official-channel access by model, region, concurrency, and usage needs. Where alternate routes remain available, reserved resources and route-level failover can reduce disruption, backed by volume pricing and end-to-end support.
Company-to-company procurement: CostRouter enters into formal service agreements with enterprise customers and accepts B2B bank transfers between corporate accounts.
Model integrity, with no silent substitution
Delivery routes may differ; the model you request does not. CostRouter never substitutes a smaller, lower-cost, or different model.
How CostRouter Routes Each Request
CostRouter receives a compatible API request, checks available model routes, and selects a route based on price, latency, availability, and your routing settings. It then returns a compatible response and records usage.
Receive the request
Your application sends a compatible request with a CostRouter API Key. The Model ID, request body, and client workflow follow the API format you already use.
Match available routes
CostRouter checks the requested model, endpoint capability, account configuration, and live channel status to find routes that can handle the request.
Select a suitable route
Among compatible, available routes, CostRouter evaluates price, latency, and availability. A lower-cost route is prioritized only when it meets the applicable reliability requirements.
Return and record
The response is returned in a compatible format, while request status, token usage, and cost are recorded for usage logs, billing review, and later optimization.
Built for Production API Workloads
CostRouter keeps routing, billing, and access rules visible so cost optimization stays auditable and controllable.
Traceable usage and billing
Review request logs, token usage, final charges, refunds, and billing records from the dashboard.
API Key-level controls
Set spend limits, expiration, model restrictions, IP allowlists, and routing group rules for each API Key.
Route quality safeguards
Routing decisions consider cost, latency, and availability, with account controls for advanced routing behavior.
Clear operating limits
Review supported models, live prices, route availability, and regional restrictions before production use.
Frequently Asked Questions
Answers for pricing, compatibility, model routing, partner supply and production use.
Is CostRouter compatible with OpenAI SDKs?
Yes. OpenAI-compatible routes use familiar OpenAI request formats. In most cases, point your existing SDK or HTTP client to the CostRouter Base URL and use your CostRouter API Key.
Why are Value groups cheaper, and are the models genuine?
CostRouter lowers inference costs through volume purchasing, preferential partner rates, reserved provider resources, multi-channel routing, and a leaner platform margin. Every route in a Value group is verified for compatibility, model authenticity, and operational reliability. Savings come from more efficient sourcing, never model degradation or substitution; groups differ mainly in delivery route, latency, stability, and available throughput. Enterprise customers can contact sales for tailored official channels, concurrency planning, volume pricing, and end-to-end support.
Do I need separate keys for GPT, Claude, Gemini and DeepSeek?
No. A single CostRouter API Key can access multiple supported models. Requests use the API format supported by each model provider, subject to model capabilities, account permissions, and route availability.
Can I choose a specific model?
Yes. You can request a specific Model ID, or use routing rules that balance price, speed, and availability for your workload.
Where can I see all supported models?
Open Models to browse current model availability, provider details, endpoints, groups and pricing information.
Is CostRouter only for developers?
No. It is useful for indie builders, startups, enterprise teams, agent workflows, and partners that provide reliable model API access.
Can channel partners list their model API routes?
Yes. Channel partners that pass CostRouter's quality review can list supported models, route availability, pricing, concurrency limits, and other service conditions.
Will I need to rewrite my application?
Most SDK or HTTP integrations only need configuration changes: update the Base URL and use a CostRouter API Key. More complex workflows can keep existing tools while gradually adopting routing controls.