Features Docs Blog Company GitHub
Try Forge →
AI assistant

`sc assistant` — AI-powered onboarding

`sc assistant dev setup` scaffolds your service config (Dockerfile, docker-compose.yaml, client.yaml). `sc assistant devops setup` walks through provisioning the parent stack. Plus an MCP server so other tools can drive SC programmatically.

Scaffold a service
Analyzes your repo and drafts Dockerfile, docker-compose, and client.yaml tailored to it. Review, tweak, deploy.
Wizard for the parent stack
Interactive prompts pick templates + resources + CI/CD config based on team size and target clouds.
Repo-aware chat
Free-form conversation about your repo — explain configs, suggest improvements, troubleshoot deploys.
MCP server built-in
Other agents (Claude Code, Cursor, Forge) connect via Model Context Protocol and drive SC: search docs, generate config, deploy with explicit user approval.

Why an assistant ships in the CLI

Most infra tools optimize for the engineer who already knows the tool. SC optimizes for the first 30 minutes — the time between “I want to deploy this” and “it’s on api.my-org.com with SSL and a database hooked up”.

sc assistant dev setup reads your project, asks two or three clarifying questions, and writes the config. The output isn’t magic — it’s the same client.yaml you’d write yourself, just produced faster and with fewer typos.

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14

# In a repo with a Go service
$ sc assistant dev setup

→ Detected: Go service, single binary, port 8080, Postgres dep
→ Suggested target: aws-lambda (single-image) + aws-rds-postgres
  Alternatives: cloudrun + gcp-cloudsql-postgres, ecs-fargate + rds
→ Domain to use? api.my-org.com
→ Writing .sc/stacks/my-svc/client.yaml ...
→ Writing Dockerfile ...
→ Writing docker-compose.yaml (for local dev) ...

Done. Review the generated files, then:
  sc secrets init --generate
  sc deploy -s my-svc -e prod

Built-in MCP integration

sc assistant mcp --port 7331 starts a Model Context Protocol server. Any MCP-compatible client — Claude Code, Cursor’s agent, Continue, the Anthropic API directly — can connect and use SC primitives as tools:

  • search_docs — look up SC features
  • generate_config — produce a client.yaml from a description
  • deploy — wired to sc deploy with the right safety checks
  • secrets_reveal / secrets_add — gated behind explicit user approval

This is what makes Forge — our flagship multi-agent product — work: Forge’s agents talk to SC over MCP for the infra layer. The agent never has direct shell access; it goes through scoped, typed tool calls.

What it isn’t

  • Not a local-dev runtime. SC isn’t trying to replace Docker Compose; it generates a docker-compose.yaml for local use and gets out of the way.
  • Not a chatbot for production debugging. sc assistant chat is configuration-aware, not metrics-aware. For runtime debugging you still want Grafana / Sentry.
  • Not free. The assistant calls a Claude API key you provide. SC ships the orchestration; the LLM bill is yours.

Ship your infrastructure on autopilot.

Try the SC parent stack in your own repo, or see what we built on top of it — Forge, our flagship AI workflow engine for teams.

Get started on GitHub Try Forge →