Deployment

Agent Smith supports four deployment models, from a single binary on your laptop to a full Kubernetes cluster with chat integration.

Deployment Options

Option	Best for	Requirements
Single Binary	CI/CD pipelines, local dev, quick scans	None (self-contained)
Docker	Persistent server, full stack with tools	Docker Engine
Kubernetes	Production, multi-tenant, auto-scaling	K8s cluster, Redis
Chat Gateway	Team collaboration via Slack/Teams	K8s or Docker, Redis

Quick Decision Guide

Do you need chat integration (Slack/Teams)?
├── Yes → Chat Gateway (Dispatcher + K8s/Docker)
└── No
    ├── Running in CI/CD? → Single Binary
    ├── Need tool containers (Nuclei/Spectral)? → Docker
    └── Production with webhooks? → Docker or Kubernetes

Start Simple

Most users start with the single binary for local testing, then move to Docker Compose for team use. Add Kubernetes and the Dispatcher only when you need chat-driven workflows or multi-tenant isolation.

Common Configuration

All deployment modes read the same YAML configuration. The discovery order is:

1. --config CLI flag (explicit path)
2. .agentsmith/agentsmith.yml (project-local)
3. config/agentsmith.yml (working directory)
4. ~/.agentsmith/agentsmith.yml (user home)

See Configuration for the full reference.

API Keys

Every deployment needs at least one AI provider key:

Variable	Provider	Required
ANTHROPIC_API_KEY	Claude	Recommended (default)
OPENAI_API_KEY	GPT-4	Optional
GEMINI_API_KEY	Gemini	Optional
GITHUB_TOKEN	GitHub repos/issues	For GitHub projects
AZURE_DEVOPS_TOKEN	Azure DevOps	For Azure projects
GITLAB_TOKEN	GitLab	For GitLab projects