Built and run by team.blue
Your own AI platform, running in Europe.
Build RAG, agents and flows, call them through an OpenAI-compatible API, and decide exactly which models may see your data. Every customer gets an isolated tensic instance on team.blue infrastructure.
- Your own isolated instance
- Hosted in Europe [countries]
- OpenAI-compatible API
Sovereignty
Sovereign, and you can check it.
Sovereign is easy to claim. Here is what it means on tensic.
Your own instance
Every customer gets a separate tensic instance with its own databases and its own address. Your data is never mixed with another customer's.
Hosted in Europe
Your instance and its data run on team.blue infrastructure in [countries or data centres].
One switch for external models
Turn off external models and tensic refuses any model that would send your data to an outside provider, embeddings and rerankers included. Unknown model types are blocked by default.
Open-weight models in Europe
Run open-weight models such as Qwen, GLM, DeepSeek and Gemma, served in Europe [confirm], or add frontier models when you choose.
Locked down by default
Credentials encrypted at rest, sign-in with SSO, OAuth or LDAPS, two-factor authentication, and roles for who can use which project and model.
API
Change one line. Keep the rest of your code.
Every tensic project is an OpenAI-compatible endpoint. Point your existing SDK at it, and every call gets the project's budget, rate limit and logging.
Illustrative instance address. [confirm: the URL customers call]
Chat with any model the project allows, streamed or not, with tools and vision.
Vectors for search, computed only by models you allowed.
Speech to text, with the same key.
Images, where the project has a generator.
Also rerank, moderations and models. Your SDKs, gateways and libraries keep working.
Build
From a prompt to a production agent, in one place.
Try it in the workbench, then call the same project from your app.
Answer from your documents
Upload files or sync them from URLs and S3, add reranking, and test answers in the workbench before you ship. Connect a MySQL or PostgreSQL database and ask it questions in plain language.
Agents that use tools
Give an agent built-in tools or any MCP server. Code it writes runs in an isolated sandbox.
Route without an LLM bill
Draw the flow: classify a request, branch, and hand it to the right project. Classification runs zero-shot, with no LLM cost.
One endpoint per project
Serve chat, embeddings, speech, images and reranking through the OpenAI API, using only the models you allow.
Works with your AI tools
Agents call external MCP servers, and tensic is an MCP server itself: connect Claude, Cursor or any MCP client to your projects.
Put it on your site
Embed a project as a chat widget with one script tag. The assistant on this page runs on tensic.
Run
Run AI like the rest of your infrastructure.
Know who uses what and what it costs, and stop it before it surprises you.
Access
Teams, roles and single sign-on. Decide who can use which projects and models.
Spend
Budgets and rate limits per project. Every call is logged with its tokens, latency and cost.
Any model
Mix open-weight and frontier models, local or hosted, without changing your app.
Your brand
Put your own name and look on the console, and turn a working project into a setup other teams or customers can reuse.
Live demo
Nothing hidden. Watch every call.
Run a demo and see, beside the answer, every call it made: which project, which model, the prompt as sent, tokens in and out, latency and cost.
| MODEL | TOKENS IN → OUT | LATENCY | COST |
|---|---|---|---|
| glm-5.3 | 182 → 412 | 1.9 s | €0.0006 |
| qwen3.8-27b | 182 → 377 | 2.4 s | €0.0005 |
| deepseek-v4-flash | 182 → 455 | 1.2 s | €0.0004 |
| gemma-4-31b | 182 → 298 | 2.0 s | €0.0003 |
Who it's for
For teams that build with AI, and for those who sell it.
Who's behind tensic
Built and run by team.blue.
tensic is built and operated by team.blue, [one line with real figures: customers, countries, years in hosting].
We run tensic the way we run hosting: on our own European infrastructure, with our own operations team. [confirm]
FAQ
Before you join.
Where is my data stored?
In your own tensic instance, on team.blue infrastructure in [confirm]. Prompts, documents and embeddings stay there unless you allow an external model.
Which models can I use?
Open-weight models served in Europe, plus any provider you connect: OpenAI, Anthropic, Gemini, Grok, Azure, Bedrock, Ollama, vLLM or any OpenAI-compatible endpoint.
Can I block external providers?
Yes. Turn off external models and tensic refuses any model that would send data outside, embeddings and rerankers included.
Do I have to rewrite my app?
No. Each project speaks the OpenAI API: change the base URL and the key.
Can I resell tensic?
Yes. Each of your customers gets an isolated instance under your brand, with plans and limits from your billing system.
When can I start?
Contact us and we will get you started.
Contact us.
Tell us what you want to build.