nosia

Nosia is a self-hosted AI RAG and MCP-enabled platform that lets you run AI models on your own data with privacy and control. It exposes an OpenAI-compatible API surface and integrates MCP to connect AI models with external tools, services, and data sources. With MCP, you can augment AI responses by invoking external tools, workflows, or data sources in a structured way while keeping data on your own infrastructure. Nosia also supports real-time streaming of AI responses via server-sent events and provides multi-format document handling (PDFs, text, websites, and QA pairs) with semantic search powered by pgvector. Use Nosia to deploy multi-tenant AI assistants that securely access your data and external tools, while maintaining compatibility with existing OpenAI-style clients.

To use Nosia and its MCP capabilities, start the Docker-based deployment and access the web/API surface. The MCP integration enables you to connect external tools or services to the AI model through MCP endpoints; you can configure tools, data connectors, and retrieval pipelines in the deployment’s environment and configuration. The platform will provide an OpenAI-compatible API surface for issuing prompts, retrieving responses, and streaming results, while the MCP layer coordinates context from connected tools and data sources during generation.

Prerequisites:

• Docker and Docker Compose installed on your host
• Internet access to pull images and fetch dependencies
• Sudo/root access if required by your OS

One-command installation (recommended):

1. Run the official installation script provided by Nosia (One-command installation pulls Docker images and config):

curl -fsSL https://get.nosia.ai | sh

2. Wait for the script to install Docker/Docker Compose (if not already present), download Nosia configuration, generate a secure .env file, and pull the required images as shown during the script output.
3. Start Nosia:

docker compose up

4. Optionally run in the background:

docker compose up -d

5. Access Nosia at:

• Web Interface: https://nosia.localhost
• API Endpoint: https://nosia.localhost/v1

Custom/Advanced installation (Docker Compose):
1) Create a directory for Nosia and navigate into it.
2) Obtain the official docker-compose files from Nosia (usually provided in the repo or guides) and place them in your directory.
3) Copy the sample .env file to your working directory and customize environment variables as needed (e.g., LLM model, embedding model, ports, and secret keys).
4) Start Nosia:
```bash
docker compose up -d

5. Verify that containers are healthy and accessible at the configured URLs.

Prerequisites recap:

• Docker and Docker Compose installed
• Suitable environment variables configured (e.g., LLM model, embedding model, secrets)
• Sufficient system resources (CPU, RAM, and storage) for the selected models and document index

Tips and common considerations:

• Nosia uses Docker Compose for deployment; the MCP integration runs as part of the Nosia services and exposes an OpenAI-compatible API surface.
• For production deployments, configure proper SSL certificates and a reverse proxy as documented in the Deployment Guide.
• If you enable augmented context or Docling-enabled processing, ensure related services are accessible and have appropriate network configurations.
• When customizing models (LLM or embedding models), ensure the corresponding environment variables (e.g., LLM_MODEL, EMBEDDING_MODEL) are set in your .env file before starting Nosia.
• Monitor logs with docker compose logs to troubleshoot startup issues or MCP wiring problems.
• For multi-tenancy, configure per-tenant data isolation in the environment and ensure credentials and access controls reflect your security requirements.