Mycelis
Mycelis
Commercial models
Start without any infrastructure: activate a commercial model with your own API key, or let Mycelis manage the key for you. Everything is reachable through the same endpoint as your hosted models.
No base fee · No minimum term · Pause any time
Two ways to activate a model
What you get
No provider accounts to juggle, no separate SDKs. One workspace, one endpoint, one place for usage and costs.
Bring keys from OpenAI, Anthropic or Google and keep the direct relationship with the provider.
No provider account needed. Mycelis manages the key; you use the model and pay per token.
Call every activated model with the same SDK and the same base URL.
Move between your own and a managed key without touching application code.
Route routine requests to a hosted open-source model and escalate to a commercial model only when needed.
Keep sensitive requests off external providers if your policy requires it.
How it works
01
Pick a commercial model from the catalog.
02
Add your own key — or select a managed key.
03
Use the OpenAI SDK with the Mycelis base URL and your Mycelis key.
Business model
Commercial models need no GPU, so there is no runtime to pay for — only usage.
Your own key
You pay the provider directly for the tokens you use.
Managed key
Usage is billed per token through your Mycelis balance.
No base fee, no minimum term
Nothing to pay while you are not using a model.
from openai import OpenAI
client = OpenAI(
base_url="https://api.mycelis.ai/proxy/v1",
api_key="pat_..."
)
response = client.chat.completions.create(
model="model-from-the-catalog", # any activated commercial model
messages=[{"role": "user", "content": "Hello"}]
)
print(response.choices[0].message.content)Common questions
With BYOK you use your own key and pay the provider directly. With a managed key Mycelis provides the access and bills the tokens you use.
Yes. Switching the key type is a setting in your workspace, not a migration.
Requests go to the provider of the model you called. To keep sensitive requests off external providers, use the personal-data filter and route them to a hosted model.
No. Commercial models run at the provider. You only need a GPU if you also want to host an open-source model.
Ready to get started?
No infrastructure, no base fee. Add hosted models whenever it pays off.