Why companies look for an alternative to Kosmoy
From Kosmoy’s own pricing and documentation, checked 9 November 2026:
- Self-hosted only: Kosmoy runs single-tenant in your Kubernetes and offers no managed service, so your team installs, upgrades and operates it.
- Price on proposal: there is no published list price, and model and cloud consumption is billed separately by each provider.
- You bring every model provider: the AI gateway routes to providers you connect, such as OpenAI, Anthropic, Azure OpenAI and AWS Bedrock, each under its own account.
- Sub-processors on request: Kosmoy's data processing addendum makes the sub-processor list available on request and lets Kosmoy keep customer personal data for up to 180 days after termination.
What Standard Compute does differently
A single API gives you frontier and open-source models, and it routes each step to the cheapest model that can do it and the cheapest provider fast enough to serve it.
Kosmoy's router also picks the cheapest model that fits a task. With Standard Compute the routing, the models and the providers come as one managed service, billed as a flat monthly plan per person or app in place of separate provider invoices.
Control is set once for the company:
Processing in Europe
Set a Europe rule for your company, a team or a single key, and every request, including retries and fallbacks, is served only by approved European inference endpoints. If no European endpoint can serve it, the request is refused before your prompt is sent. Our EU gateway terminates TLS and runs in Finland, and your account data is stored in Frankfurt.
Zero data retention
Turn on zero data retention for your company, a team or a key, and every request is routed only to inference endpoints that commit to keeping no copy of your data. Teams and keys inherit the requirement and can never switch it off underneath you.
No training
Standard Compute never trains on your prompts or outputs. With a region rule or zero retention turned on, every request goes only to providers that commit not to train on it.
Single sign-on
Connect Microsoft Entra ID or any SAML identity provider so your people sign in with their work accounts. Employees sign in to Claude Desktop and Claude Code with their Microsoft work account through our EU gateway, and access follows your Entra groups through SCIM provisioning.
Teams and keys
Organize your company into teams, invite colleagues by email, and give every person and every app its own key, all under one company account.
Budgets
Every person and app gets a fixed monthly budget that never overruns, and you can set spending caps per key and for the whole company. Optional top-off from a prepaid company balance keeps critical agents running.
Allow-lists
Choose exactly which inference providers may handle your data, for the whole company, a team or a single key. New providers are never added without an admin's approval, and fallbacks can never route around your list. Decide which models each key can use.
Audit
Every admin change is recorded with before and after values, and you can export a record of every request showing the model, the provider that served it and the region rule in force, kept for 12 months.
Contract
Sign our Data Processing Agreement in the dashboard. It commits us to no training on your data, no retention of your content, and breach notification within 72 hours. We publish every sub-processor and give you 14 days' notice before adding a new one.
Standard Compute and Kosmoy compared
| Standard Compute | Kosmoy | |
|---|---|---|
| Pricing | Flat monthly plan per person or app with no per-token surprises; Stripe invoices with your company name and VAT number | Per proposal; model and cloud consumption billed by each provider |
| Cost routing | Each step goes to the cheapest model that can do it | Router picks the cheapest model that fits; fallback and load balancing |
| Models | Claude, GPT, Gemini and open models (GLM, DeepSeek, MiniMax, Kimi) through one API | Providers you connect: OpenAI, Anthropic, Google, Mistral, Azure, Bedrock, vLLM, Ollama |
| Processing in Europe | Europe rule per company, team or key; retries and fallbacks stay on approved European endpoints; refused before sending if none qualifies | Runs where you deploy it; data Kosmoy holds may be processed where Kosmoy or its sub-processors have facilities |
| Zero data retention | Per company, team or key; lower levels cannot switch it off | Logs stay in your infrastructure; provider retention set with each provider |
| Training on your data | Never trains on prompts or outputs | Not published |
| Single sign-on | SAML SSO, Microsoft Entra ID included; Claude Desktop and Claude Code sign-in with a Microsoft work account, SCIM from Entra | OIDC and SAML, tested with Okta, Microsoft Entra ID and Auth0; SCIM not published |
| Teams and keys | Teams, roles, a key per person and per app | Role-based access for users, teams, applications and agents |
| Spend limits | Fixed monthly budget per key; requests are refused once it is used; caps per key and per company | Limits by model, app, team, environment and use case |
| Provider and model allow-lists | Provider allow-list per company, team or key; new providers need an admin; models chosen per key | Model approval; guardrails for PII, toxic language and prompt injection |
| Audit log | Admin log with before and after values; request export with model, provider and region rule, kept 12 months | Guardrail decisions, approvals and overrides logged |
| DPA and sub-processors | DPA signed in the dashboard; public sub-processor list with 14 days' notice | DPA published; sub-processor list on request |
| What you operate | Managed service; nothing to host | Self-hosted in your Kubernetes; no managed service |
When Kosmoy fits
Kosmoy fits a large regulated enterprise with a platform team that wants the AI control plane inside its own Kubernetes, with guardrails for PII and prompt injection and evidence mapped to the EU AI Act.
Moving from Kosmoy to Standard Compute
Standard Compute’s API runs on stdcmpt.com. https://eu.stdcmpt.com is our EU gateway in Finland.
- Create a company workspace and set the rules once for the company: a Europe rule, zero data retention, and the providers and models you approve. Teams and keys inherit them.
- Create a key for each person and each application, each with its own monthly budget.
- Point applications at
https://eu.stdcmpt.com/v1in place of your Kosmoy deployment's base URL, with a Standard Compute key. OpenAI-compatible client code stays the same. - Recreate model approvals as model selections per key and the company provider allow-list; budgets move to a fixed monthly budget per key.
- Point Claude Code at
https://eu.stdcmpt.com. - Connect Microsoft Entra ID or your SAML identity provider so your people sign in with their work accounts, and set up Claude Desktop and Claude Code sign-in with Microsoft work accounts.
from openai import OpenAI
client = OpenAI(
base_url="https://eu.stdcmpt.com/v1",
api_key=os.environ["STANDARD_COMPUTE_API_KEY"],
)
# "standardcompute" sends each step to the cheapest
# model that can do it. Or pin a model:
# model="anthropic/claude-sonnet-5.5"
client.chat.completions.create(
model="standardcompute",
messages=messages,
)# Claude Code adds /v1/messages itself
export ANTHROPIC_BASE_URL=https://eu.stdcmpt.com
export ANTHROPIC_AUTH_TOKEN=YOUR_STANDARD_COMPUTE_KEY
claudeQuestions
Is there a managed alternative to Kosmoy?
Standard Compute is a managed gateway with nothing to host. Teams, a budget per key, provider and model allow-lists, Microsoft Entra SSO and an audit log are set self-serve in the dashboard.
Do we need our own contracts with OpenAI, Anthropic and Google?
No. A single Standard Compute API gives you frontier and open-source models, on flat monthly plans per person or app with Stripe invoices carrying your company name and VAT number.
Can we restrict which providers handle our data?
Yes. Choose exactly which inference providers may handle your data, for the whole company, a team or a single key. Lower levels can only narrow the rules, new providers are never added without an admin's approval, and fallbacks can never route around your list.
What does the audit log record?
Every admin change is recorded with before and after values, and you can export a record of every request showing the model, the provider that served it and the region rule in force, kept for 12 months.
How do we migrate from Kosmoy?
Create a company workspace and set your rules, then point your applications at https://eu.stdcmpt.com/v1 with a Standard Compute key. OpenAI-compatible client code stays the same.
Set up your company
Create a company workspace, invite your team and set your rules from the dashboard. Security questionnaires and procurement questions go to contact@standardcompute.com.
Create a company workspaceor write to contact@standardcompute.com
Sources (checked 9 November 2026)
- Kosmoy homekosmoy.com
- Product factskosmoy.com
- Pricingkosmoy.com
- AI gatewaykosmoy.com
- Platformwww.kosmoy.com
- Data processing addendumkosmoy.com