The server, the gateway and the keys: an OpenAI- and Anthropic-compatible AI running inside your walls.
The Artemia API is not a subscription to a distant model: it is a machine and a gateway installed inside your walls. Your models run on your hardware, behind your network, and answer to both market standards — OpenAI and Anthropic. Your developers keep their tools, your teams build their own automations, your agents run locally. No line of code, no document, no prompt ever leaves your infrastructure.
Why on-premise AI? Security, Cloud Act, performance, costs: our guide to local AI for business.We do not sell access to a standardised API, but an artificial intelligence infrastructure designed for your use cases and handed over in working order. Every project is run end to end: performance, security, cost control and integration with what you already have.
The generative AI model landscape moves every quarter. Our monitoring is continuous: when a stronger model ships in a category you rely on, we qualify it, validate it against your use cases and make it available on your instance.
A single gateway, as many contexts as you have teams. Developers, support functions and business units plug in their own tools — and traffic never leaves the network.
Chat, edits and fill-in-the-middle completion right inside the IDE, served by a code-specialised model. The repository stays on the workstation and on the server: nothing is sent to a third-party vendor.
The gateway exposes Anthropic's Messages API. Two environment variables are enough for Claude Code, or any client of the anthropic SDK, to work against your instance — tools, streaming and thinking blocks included.
n8n, Make or your own scripts call the API exactly as they would call OpenAI. You build complete chains — read, decide, write back into your tools — with no external dependency and no meter running.
Open-source agent harnesses point at the same URL. They read your files, call your systems and run long task chains on a model that has no internet access of its own.
The embeddings endpoint vectorises your document bases for search, classification and RAG. Both the vectors and the source documents stay in your storage.
Vision models read an invoice, a drawing or a field photo and return structured data, ready to feed your ERP or your document management system.
No proprietary SDK, no in-house format: the gateway answers to both market standards. Here are the four configurations we are asked for most — the others fit on the same single line.
# The OpenAI standard, served by your own instance
curl -s https://ai.your-company.com/v1/chat/completions \
-H "Authorization: Bearer $ARTEMIA_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "Mia-Agent",
"messages": [
{"role": "user", "content": "Summarise these minutes."}
]
}'The OpenAI response schema, streaming included — and an X-Request-Id on every response.
import os
from openai import OpenAI
client = OpenAI(
base_url="https://ai.your-company.com/v1",
api_key=os.environ["ARTEMIA_API_KEY"],
)
reply = client.chat.completions.create(
model="Mia-Agent",
messages=[{"role": "user", "content": "Hello!"}],
)The official OpenAI SDK, unchanged. The anthropic SDK works exactly the same way.
{
"env": {
"ANTHROPIC_BASE_URL": "https://ai.your-company.com",
"ANTHROPIC_AUTH_TOKEN": "your-artemia-key",
"ANTHROPIC_MODEL": "Mia-Agent",
"ANTHROPIC_DEFAULT_OPUS_MODEL": "Mia-Coder",
"ANTHROPIC_DEFAULT_HAIKU_MODEL": "Mia-Vision"
}
}Claude Code then runs on your models, without reaching a single external server.
# The n8n OpenAI node, simply repointed
Base URL : https://ai.your-company.com/v1
API Key : your Artemia key
# The Chat, Embeddings and AI Agent nodes
# then list your own models.Same principle for Make, Dify or any OpenAI-compatible VS Code extension.
The catalogue is declarative: GET /v1/models returns the exposed models and their capabilities. Your integrations read the list instead of hard-coding identifiers, and a newly added model shows up without a redeploy.
This catalogue is an example: the models exposed on your instance are chosen with you, based on your use cases and the hardware installed.
A gateway is worth nothing without the uses it powers. We ship automation chains that are already proven and ready to plug into your tools — and we build the ones that do not exist yet with you.
Invoices, purchase orders, contracts: the useful fields are read and returned as structured data to your ERP.
Your reference material vectorised and queried in plain language, with answers that cite their sources.
Incoming mail classified, routed to the right team, with pre-filled reply drafts.
Transcription, summary and decision log dropped straight into your workspace.
Internal and public sources reviewed every morning, with a summary sent to the teams concerned.
Merge requests analysed, fixes suggested and unit tests generated.
Every organisation has its own processes, tools and exceptions. We start from your real use cases, prototype with the teams concerned, then ship the automation integrated into your environment.
Talk through your use caseGeneric APIs let you start in minutes; they show their limits as soon as usage becomes daily and industrial. At scale the maths change — on the invoice, on confidentiality, and on how dependent you are.
) Tell us what your teams do today, with which tools and at what volume. We size the machine, choose the models and deliver the access — with a technical demo to back it up.
A question first? Read the API FAQ · info@artemia.ai