Building your first AI agent shouldn’t require a week of infrastructure work. EnclaveAI is designed so you can go from zero to a deployed, monitored agent in an afternoon. Here’s how.
What is an Enclave?
An Enclave is your isolated AI environment — a self-contained space where your agents, data sources, tools, and configuration all live together. Think of it as a secure workspace for your AI: everything it needs is inside, nothing leaks out.
Each Enclave runs on EU infrastructure, has its own audit log, and is fully isolated from other customers. You own what runs inside it.
Step 1: Create your Enclave
Log in and click New Enclave. Give it a name — something that matches your project or team. You can create as many Enclaves as you like: one per project, one per environment, or one per client.
Once created, your Enclave gets a unique API endpoint. That’s the address your application will use to talk to your agents.
Step 2: Connect your data
Agents are only useful if they know about your business. In the Data Sources tab, connect the information your agents need:
- Upload documents (PDFs, text files, spreadsheets)
- Connect a database or vector store
- Link to external APIs
EnclaveAI indexes your data and makes it available to agents at query time. Your data never leaves the Enclave.
Step 3: Configure your first agent
Open the Agents tab and click New Agent. You’ll be asked to:
- Choose a base model from the supported model list
- Write a system prompt describing what the agent should do
- Attach the data sources and tools the agent can access
- Set any output constraints or guardrails
If you prefer to work in code, you can define the entire agent in a config file using the SDK and push it to your Enclave via the CLI.
Step 4: Test it
Use the built-in chat interface to test your agent before deploying. Try a few queries, check that it’s using the right data, and verify the responses look right.
Every test run is logged — you can review them in the Run History tab at any time.
Step 5: Integrate via API
When you’re ready to go live, call your Enclave’s API endpoint from your application:
curl -X POST https://your-enclave.enclaveai.eu/v1/chat \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{"message": "What are our refund policies?"}'Responses are streamed by default. Context from previous messages in a session is handled automatically.
What happens next
Once your agent is live, EnclaveAI keeps working in the background. The built-in QA engine monitors response quality and flags issues. Run history gives you a full trace of every query. And when you want to improve things — swap the model, update the data, tweak the prompt — you do it in the Enclave and the changes take effect immediately.
That’s it. One afternoon, one Enclave, one agent in production.