Host AI Agents

Back to home

Runtime and language support

Upsun supports multiple runtimes through container-based deployments including Python, Node.js, PHP, Ruby, Go, Java, and more. For the complete list of supported runtimes and their versions, see the [runtime types reference] (/create-apps/app-reference/single-runtime-image.md#types).

Configure your runtime in the .upsun/config.yaml file. The platform builds your application in a container with the specified runtime and dependencies.

LLM API integration

You can integrate your AI agent with any LLM API that your chosen runtime supports:

OpenAI API: Use the official OpenAI client libraries for Python, Node.js, and other languages
Anthropic Claude API: Use the Anthropic client libraries for Python, Node.js, and other languages
Google Gemini: Use the Google AI client libraries for Python, Node.js, and other languages
Azure OpenAI: Use the Azure OpenAI client libraries for Python, Node.js, and other languages
AWS Bedrock: Use the AWS Bedrock client libraries for Python, Node.js, and other languages
Other providers: Integrate with any API that provides HTTP endpoints

Your application code handles the API calls and response processing. Upsun doesn’t restrict which services you can use.

Environment management

Upsun provides isolated environments for development, testing, and production:

Branch-based environments: Each Git branch creates a separate environment
Data isolation: Each environment has its own services and data
Easy cloning: Clone production data to development environments for testing
Environment variables: Store API keys and configuration securely using variables

Tutorials

Deploy a RAG-based conversational agent with Chainlit: Build a Retrieval-Augmented Generation agent using Chainlit, llama_index, and OpenAI, then deploy it on Upsun. See the Chainlit deployment tutorial.
Access Documentation contextually via Context7 + MCP: Use the Model Context Protocol to let AI assistants fetch your Upsun documentation in real-time. See the Context7 MCP article.
Use the Upsun API to automate Agent deployment: Automate deployments, environment management, and configuration through the Upsun API. See the API usage guide.

Configuration example

Here’s a basic configuration for a Python AI agent. For more configuration options, see the complete application reference:

# .upsun/config.yaml
applications:
  ai-agent:
    source:
      root: "/"
    type: "python:3.11"
    mounts:
      ".data": source: "storage", source_path: "data"
    web:
      commands:
        start: "python agent.py"
      upstream:
        socket_family: tcp
        locations:
          "/": root: "", passthru: true

Application Code

For examples of how to implement AI agents with different frameworks and APIs, see the AI and Machine Learning tutorials on DevCenter.

Deploy your Agent

Add your code to Git:

git add .
git commit -m "Add AI agent service"

Set your OpenAI API key as an environment variable using the [CLI] (/administration/cli/_index.md):
```
upsun variable:create env:OPENAI_API_KEY --value=<your_key>
```
For more information about setting variables, see [how to set variables] (/development/variables/set-variables.md).
Deploy to Upsun:
```
upsun push
```

Your agent will be available at your Upsun environment URL. The platform handles the containerization, networking, and scaling automatically. For more deployment options, see deploy your project.

Key benefits

Runtime flexibility: Choose the programming language and version that fits your needs
Service independence: Use any LLM API or external service
Environment isolation: Test changes safely in separate environments
Automated deployment: Deploy through Git pushes or API calls
Scalability: The platform handles load balancing and resource allocation

For more information about building and deploying applications, see the configure apps section.

Integrated environments

Supported environments