Agents

Agents are the autonomous workers in the Red Station platform. Unlike generic "chatbots" that simply output text, Red Station agents are specialized, role-based entities that can plan, execute, and use tools.

The Agent Runtime

Agents operate within a secure runtime that provides:

• Memory: Short-term conversation history and long-term vector recall.
• Tools: Access to specific capabilities (e.g., "Read File", "Run Test").
• Policy: Strict enforcement of what an agent can and cannot do.
• Identity: Each agent has a specific role (Planner, Coder, Reviewer).

Built-in Agents

Red Station ships with several optimized agents out of the box:

ID	Role	Description
planner	Architect	Analyzes requirements and creates implementation plans.
coder	Developer	Writes, modifies, and refactors code.
reviewer	QA	Reviews code changes against best practices.
security	Security	Scans for vulnerabilities (Pro/Enterprise).

CLI Commands

Manage agents directly from the terminal:

# List all available agents
pilot agent list --verbose

# Show detailed configuration for an agent
pilot agent show planner

# Run an agent ad-hoc on a task
pilot agent run coder --task "Refactor the login logic in auth.py"

# Create a new custom agent
pilot agent create my-custom-agent

Configuration (YAML)

Agents are defined in YAML files located in ~/.pilot/agents/ or .pilot/agents/.

agent:
  id: infra-agent
  name: Infrastructure Agent
  version: 1.0.0
  description: Manages Terraform and Kubernetes resources.
  category: infrastructure

model:
  provider: ollama
  name: codellama:13b
  temperature: 0.2

system_prompt: |
  You are an expert DevOps engineer...

tools:
  - type: mcp
    server: filesystem
    allowed: [read_file, write_file]
  - type: command
    name: terraform

behavior:
  autonomy: supervised

See Agent YAML Spec for full details.