Skip to content

Factory AI Droid CLI

What it is

An AI-powered coding agent that works directly within your project to perform tasks like code reviews, security scans, and feature implementation. Droid leverages LLMs (typically Claude) to interact with your codebase and supports specialized "droids" (sub-agents) for specific domains like infrastructure, security, or frontend development. It is designed to be a "knowledge-aware" agent that understands project-specific context through a droid.yml configuration.

What problem it solves

Automates repetitive development tasks such as code review, security scanning, and feature implementation by delegating them to specialized AI sub-agents. It solves the "blank page" problem for new features and the "tedium" problem for routine maintenance.

Where it fits in the stack

Development & Ops. Acts as a CLI-based AI coding agent with domain-specific sub-agents. It sits alongside tools like Aider but focuses more on autonomous multi-step tasks.

Architecture overview

Droid uses a controller-worker architecture where the main CLI orchestrates specialized "droids" based on the task description.

flowchart TD
    A[Droid CLI] --> B{Task Router}
    B -- Infra task --> C[Terraform Droid]
    B -- Feature task --> D[Feature Droid]
    B -- Review task --> E[Review Droid]
    C & D & E --> F[LLM Reasoning]
    F --> G[File System / Git]

Getting Started (droid.yml)

Configuration is handled via a droid.yml file in the repository root:

version: 1
project:
  name: "My Awesome App"
  stack: ["python", "fastapi", "postgresql"]
droids:
  - id: "security-reviewer"
    type: "reviewer"
    rules:
      - "Ensure all API endpoints have authentication"
      - "Check for SQL injection patterns in raw queries"
  - id: "feature-builder"
    type: "coder"
    context: ["docs/api_specs/", "src/models/"]

CLI Usage Examples

  • Run a security scan: 
    droid run security-reviewer --target ./src
  • Implement a feature: 
    droid build "Add a new endpoint to list users with pagination"
  • Interactive session: 
    droid chat --context "current-file"

Typical use cases

  • Automated code reviews and security scans in CI/CD pipelines.
  • Bootstrapping new modules or features from a specification.
  • Domain-specific tasks using specialized droids (infrastructure, security, frontend).
  • Refactoring legacy code according to new style guidelines.

Strengths

  • Specialized Sub-agents: Domain-specific droids provide higher accuracy than general-purpose agents.
  • CLI-first Design: Integrates seamlessly with existing terminal workflows and CI/CD (GitHub Actions).
  • Knowledge Awareness: Uses droid.yml to maintain persistent project context.

Limitations

  • Model Dependency: Performance is tied to external LLM providers (typically Claude).
  • Complexity: Requires initial configuration for complex multi-droid workflows.
  • Ecosystem: Smaller community compared to mainstream tools like GitHub Copilot.

When to use it

  • When you want AI-driven automation for code reviews, security scans, or feature work.
  • When you need domain-specific AI agents within a CI/CD pipeline.
  • For large-scale refactorings where consistency is paramount.

When not to use it

  • For quick, single-line completions (use GitHub Copilot or Codeium).
  • When you require full local model execution (see Llama.cpp).

Sources / references

Contribution Metadata

  • Last reviewed: 2026-06-01
  • Confidence: high