Keep your SDK
Use the client your agents already run.
OpenAI SDKAnthropic SDKApertis SDKFetch clients
Model Delivery Infrastructure for AI Agents
Managed model supply for teams running AI agents. Route across 30+ providers, deliver in native OpenAI and Anthropic formats, and reconcile every call against a budget.
How Apertis delivers a model call
Supply
30+ providers, qualified route by route.
Route
Every call takes a healthy route, with rate caps and circuit breaking per model.
Deliver
Native OpenAI and Anthropic formats. Unsupported fields are refused, never silently dropped.
Reconcile
Every call lands on a budget, with its key, model, tokens and cost.
One request path for every agent loop
Budgets, access policy, routing and cost attribution sit on the one path every model call already takes.
Budgets and Spend Caps
Give every team, agent and workload its own budget. Requests stop at the cap you set, not at the invoice.
Per-key budgetsMonthly limitsKey expiry
Model Catalog
Current model IDs, prices and provider routes in one catalog, so a model review starts from one list.
500+ models30+ providers
Routing and Failover
Model choice stays explicit. Provider health and your fallback order decide the next route.
Fallback orderRoute health
Cost Attribution
Every request records its key, model, tokens and cost, so spend traces back to a team, an agent or a workload.
Per-request recordsPer-key usage
Access Policy
Scope each key to the models it may call and the networks it may call from. Anything outside is refused.
Allowed modelsIP allowlists
No Rewrite
Agents and apps keep their OpenAI or Anthropic SDK. Only the base URL changes.
OpenAI formatAnthropic format
Billing API
Read remaining credit from code, and gate a long agent run on it before it starts.
Remaining creditsProgrammatic checks
GET Use one or all.Model providers, spend policy, and agent access governed together.
Supported Providers
Logos are trademarks of their respective owners.
Deploy a governed AI layer in minutes
Keep your current SDK, issue controlled keys, set budgets and model policy, then route agent traffic through Apertis.
Set base URL
Send requests through the governed Apertis endpoint.
https://api.apertis.ai/v1/chat/completionsChoose model policy
Keep model, fallback, and credit checks explicit.
- Model
- claude-sonnet-5.5
- Controls
- budget + fallback
- Credits
- /v1
/dashboard /billing /credits
Trending Models
500+ active models on 30+ providers
Claude Opus 5.5
Flagship Claude route for long-horizon agentic coding.
Agentic codingNative Anthropic
claude-opus-5.5Claude Sonnet 5.5
Fast Claude route for iterative edits and review.
Fast agent loopsNative Anthropic
claude-sonnet-5.5GPT-6.1 Sol
Near-Astra GPT-6 route for agentic coding at lower cost.
Agentic codingNative OpenAI
gpt-6.1-solProvider coverage across direct endpoints and fallback pools.
AWS BedrockOpenAI, Claude, Grok
U.S. directPerplexity + Cohere
Mistral USExplicit US endpoint
Mistral EUDefault API hosting
BasetenUS, EU, UK, AU
Open poolDeepSeek, Qwen, Z.AI
Bedrock APACTokyo, India, Singapore
NovitaOpen-model hosting
30+providers500+models8coverage anchorsPolicyfallback + spend controls
Frequently asked questions
Short answers for teams checking SDK support, model coverage, pricing, and agent safety before routing traffic.
What is Apertis?
Apertis is model supply and delivery infrastructure for AI agents and AI apps. It gives teams one governed API path across 30+ AI model providers, plus virtual keys for issuing controlled access per team or agent, budgets and model policy to keep usage inside spending and model-access boundaries, routing and failover across providers, and drop-in integration — keep your SDK and change only the base URL.
Which AI models and providers does Apertis support?
30+ providers and hundreds of models: OpenAI, Anthropic, Google (Gemini + Vertex), AWS Bedrock, Mistral, DeepSeek, Alibaba (Qwen), Moonshot, Zhipu (GLM), SpaceXSI, Together, and Groq. Coverage includes chat, embeddings, images, speech, and audio transcription. The current model list is browsable at https://apertis.ai/models.
Which SDKs can I use with Apertis?
Apertis supports multiple SDK surfaces: OpenAI-format clients for /v1/chat/completions, /v1/responses, /v1/embeddings, /v1/images/generations, and /v1/audio/transcriptions; native Anthropic Messages at /v1/messages; direct HTTP; and the first-party Apertis SDK at https://github.com/apertis-ai/apertis-sdk for richer ergonomics over Apertis-specific features. Streaming, function calling, vision, and tool use all pass through.
How does Apertis pricing work?
Two ways to pay, mix as you like. Subscription tiers are Lite at $12/mo, Pro at $25/mo, Plus at $60/mo, and Max at $200/mo — each with a defined quota for predictable monthly cost. Pay-as-you-go lets you top up credits and pay per million tokens at per-model rates, with no minimum spend. When a subscription quota is exhausted, requests fall back to PAYG credits with a hard cap you control, so a runaway agent loop won't drain your card. Full breakdown at https://apertis.ai/subscribe.
How is Apertis different from OpenRouter / ZenMux / 9router?
Apertis uses paid multi-provider routes and does not depend on reusing consumer subscription OAuth tokens. It adds virtual keys, budgets, model policy, fallback, and request-level usage records around OpenAI-compatible and Anthropic API traffic. Model and route availability still depends on provider policy, region, account configuration, and the selected model. Apertis is operated by STIMA AI LLC; current product and setup details are available at https://docs.apertis.ai.
Is Apertis safe to use with Hermes Agent / OpenCode / Claude Code / Kilo Code / Cursor / Cline?
Hermes Agent, OpenCode, Claude Code, Kilo Code, Cursor, and Cline can connect when the installed version supports a custom OpenAI-compatible or Anthropic endpoint. Configuration and supported capabilities vary by agent version and selected model, so use the current setup guides at https://docs.apertis.ai/installation and test the required tools, streaming, and model features before production use.
Need implementation details?Open docs