Skip to main content

What is Station?

Station is a secure, self-hosted platform for managing AI agents with MCP (Model Context Protocol) tool integration. It provides:
  • AI Agent Runtime: Execute agents with any AI provider (OpenAI, Gemini, Anthropic, Ollama)
  • MCP Server: Expose agents as MCP tools for AI editors like Claude Desktop and Cursor
  • Management UI: Web-based interface for configuration and monitoring
  • Bundle System: Package and distribute agent configurations

Architecture

Key Concepts

Environments

Station organizes configurations into environments (e.g., default, production, development). Each environment has its own:
  • Agents
  • MCP server configurations
  • Variables
~/.config/station/
├── config.yaml              # Global settings
├── station.db               # SQLite database
└── environments/
    ├── default/
    │   ├── template.json    # MCP servers config
    │   ├── variables.yml    # Environment variables
    │   └── agents/
    │       └── my-agent.prompt
    └── production/
        └── ...

Agents

Agents are defined as .prompt files using dotprompt format:
---
name: code-reviewer
description: Reviews code for best practices
model: openai/gpt-4o
tools:
  - filesystem_read
  - github_pr_comments
---

You are an expert code reviewer. Analyze the provided code and suggest improvements.

MCP Servers

MCP servers provide tools to your agents. Configure them in template.json:
{
  "mcpServers": {
    "filesystem": {
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/workspace"]
    }
  }
}

Bundles

Bundles package complete environment configurations for easy distribution:
# Create a bundle from your environment
stn bundle create default --output my-bundle.tar.gz

# Install a bundle from CloudShip
stn bundle install <bundle-uuid> my-environment

Running Modes

ModeCommandUse Case
Serverstn serveLocal development, direct API access
Dockerstn upIsolated deployment, CI/CD
stdiostn stdioDirect MCP integration with editors

Next Steps