Renkler tuhaf mı görünüyor? Samsung Internet tarayıcısı koyu modda site renklerini değiştiriyor olabilir. Kapatmak için Internet menüsünden Ayarlar → Kullanışlı Özellikler → Labs → Web site koyu temasını kullan seçeneğini etkinleştirebilirsiniz.
LLM routing gateway
A single, smart entry point for your AI models.
A single OpenAI-compatible endpoint. The difficulty of every request is classified and the request is routed to the most cost-appropriate model that can do the job; requests are stopped before team budgets are exceeded, and every request is logged.
RuneGate
Smart routing
Misused models and scattered integrations leave teams unable to see what they spend, and the budget slips out of control.
A simple question and a complex analysis usually go to the same expensive model.
Every provider needs its own API key, its own integration and its own invoice.
Teams cannot see what they are spending in real time; budget overruns show up when the invoice arrives.
Trying a new model or provider means re-integrating the application.
Your application sends requests to one endpoint; RuneGate decides which model to use separately for every request.
Every application that uses an OpenAI client, n8n flows and scripts switch to RuneGate by changing only the base_url and the API key. Use rune-auto as the model name and let RuneGate choose; pin the tier or the model if you prefer.
# OpenAI client: two lines change
client = OpenAI(
base_url="https://gate.sirket.com/v1",
api_key="sk-ekip-anahtari",
)
# rune-auto: RuneGate makes the choice
r = client.chat.completions.create(
model="rune-auto",
messages=[{"role": "user",
"content": "Summarize the contract."}],
)The right model is chosen automatically by the difficulty of the request: simple jobs go to an economical model, hard jobs to a powerful one. You can pin the tier or the model yourself.
OpenAI and Google Gemini models, plus open-source models such as DeepSeek, Qwen and GLM, are used from one endpoint with one team API key.
Images, PDFs, Excel sheets and audio files are sent through the same endpoint; the request is routed to a model that can understand them.
A monthly budget for every team; when it runs out, requests are stopped before they reach the provider. Allowed tiers and a request limit are set per team, and spending is visible in real time.
Which request went to which model, and why, can be traced in the logs; the panel shows daily spending, model distribution and team usage.
Used as a managed service; for organizations whose data cannot leave, it is installed on-premise on RuneBox and routes requests to local models.
RuneGuard is an optional data protection module that runs inside RuneGate. It replaces sensitive information in a request with placeholders before it reaches the model provider, then puts the right value back in place of each placeholder in the model's answer. The user gets a complete answer, and the sensitive data never reaches the model.
On the way out, sensitive information becomes placeholders; on the way back, each placeholder is replaced with the real value.
Besides personal data such as national ID numbers, phone numbers and email addresses, sensitive information specific to your organization or field can be defined too; it is masked, or the request is blocked entirely.
RuneGuard runs as part of RuneGate; your applications and your integration stay exactly as they are.
Offered as a separate module; switched on for teams that work with sensitive data, and it makes your KVKK assessments easier.
Summarize the application: national ID 12345678901, phone 0532 000 00 00, email ayse@ornek.com; the application is part of Project Atlas.
Summarize the application: national ID [NATIONAL_ID], phone [PHONE], email [EMAIL]; the application is part of [PROJECT].
Summary: the application under [PROJECT] belongs to the customer with national ID [NATIONAL_ID]; contact via [PHONE] and [EMAIL].
Summary: the application under Project Atlas belongs to the customer with national ID 12345678901; contact via 0532 000 00 00 and ayse@ornek.com.
Simple requests stay on the economical model; the savings depend on the mix of your traffic.
Requests stop before a team budget is exceeded; who is spending what is visible at any moment.
All providers are managed from one place; adding a new model needs no change in the application.
Designed with the experience RuneLab gained in enterprise AI projects; routing decisions can be audited from the logs.
For every team that puts AI into its products and workflows and wants to manage the cost.
Managing several products and models through one integration; so that changing a model needs no deployment.
One key, automatic model selection and a budget cap for flows built with n8n and similar tools.
Per-team budgets, real-time spend and single-invoice visibility; central management of provider API keys.
Trying new models without changing code; tracing which request went to which model in the logs.
The assistant that makes your documents answerable and gives every answer with its source.
See the product →
GPU infrastructure that runs AI inside your organization, with no data leaving.
See the product →
A single endpoint for your models: routes each request to the right model and keeps the budget per team.
This pageAn LLM gateway is a single entry point that sits between your applications and language model providers. Every request passes through it, so model choice, team budgets and logs are managed in one place. RuneGate does this with one OpenAI-compatible endpoint; the only changes in your applications are the address and the API key.
No. RuneGate is OpenAI-compatible: only the address and the API key change in the client; your current usage, streaming included, keeps working as it is.
Yes. rune-auto leaves the choice to RuneGate; if you prefer, you pin the tier or the model itself. The tiers allowed for each team are set in the admin panel.
The request is completed automatically with another model; your application keeps calling the same endpoint and nothing needs to change.
It stops. When the team's budget runs out, the request is stopped before it reaches the provider and the application receives a budget-exceeded response. Budgets renew monthly; spending is tracked in real time in the panel.
OpenAI and Google Gemini models, plus open-source models such as DeepSeek, Qwen and GLM. A new model is added without any change to your applications.
Logs are visible only to your organization's authorized panel users; in an on-premise installation they stay on your network. The retention period and deletion policy are agreed together with your organization.
With the RuneGuard module, personal data and sensitive information specific to your organization are replaced with placeholders before the request is sent to the provider, and the real values are put back when the answer returns. If data must never leave your organization, RuneGate runs with local models on RuneBox.
Yes. RuneGate is installed on-premise on RuneBox and routes requests to the local models on the box; applications use the same OpenAI-compatible endpoint. It is also offered as a managed cloud service.
Savings depend on the mix of your traffic: the larger the share of simple requests, the bigger the difference. In the demo we go through routing and cost logs together using your own sample requests.
We prepare a quote for your organization based on the number of teams, the monthly request volume and the providers to be used. In the demo we can look at routing and cost records with your own sample requests.
Tell us the models you use, your monthly request volume and your team structure; we will get back to you within one business day.