Station provides a secure, self-hosted runtime for building, managing, and deploying AI sub-agents. It targets engineering teams seeking to integrate AI capabilities into their development workflows and infrastructure, offering centralized management, environment isolation, and server deployment for agents that securely access internal systems.

How It Works

Station functions as an MCP (Model Context Protocol) runtime, augmenting development environments with deployable sub-agents. Agents are defined via .prompt files, combining AI prompts with MCP tools that interface with various services (Filesystem, Docker, Kubernetes). Station facilitates a lifecycle from local development to server deployment, enabling versioned agents and tools for intelligent infrastructure, security, and automation tasks. Its core advantage lies in providing a unified framework for agent creation, management, and deployment across diverse environments.

Quick Start & Requirements

Installation is initiated via a shell script: curl -fsSL https://raw.githubusercontent.com/cloudshipai/station/main/install.sh | bash. A quick start command stn bootstrap --openai sets up OpenAI integration, a default "Hello World" agent, a Playwright agent, and a DevOps Security Bundle. A prerequisite is setting the OPENAI_API_KEY environment variable. Station also supports manual setup for advanced configurations. Documentation and guides are available.

Highlighted Details

• Station Bundles: Pre-configured, portable environment packages containing specialized AI agents, MCP tools, and workflows, installable via UI or CLI (stn bundle install). The DevOps Security Bundle includes vulnerability scanning and IaC validation tools.
• CI/CD Integration: Supports deployment patterns including Agent-as-a-Service via Docker, direct execution in CI runners, and programmatic orchestration with Dagger modules, with examples for GitHub Actions.
• Interactive Development Playground: A browser-based environment (genkit start -- stn develop --env dev) for interactive agent testing, debugging, prompt iteration, and execution flow analysis.
• Sub-Agent Definition: Agents are defined in .prompt files supporting rich input/output schemas, metadata, and execution configurations, allowing for dynamic agent behavior.

Maintenance & Community

Community support and discussions are available via Discord. Bug reports and feature requests can be submitted through the Issues tracker. A Bundle Registry hosts community agent templates.

Licensing & Compatibility

Station is licensed under AGPL-3.0, which permits free use and open-source contributions but requires derivative works to be shared under the same license. This may impose compatibility considerations for closed-source or commercial applications.

Limitations & Caveats

The project is currently in a beta release, meaning breaking changes may occur between versions; pinning to specific versions in production is recommended. Some security tools require Docker socket access. System requirements include Linux, macOS, or Windows, with minimum memory of 512MB and storage for the binary and agent data.