The SMF Works Project — Where AI Meets Humanity
← Back to getting started
General

Step 1: Pick a task

Start with one narrow, well-defined task. Good first tasks:

  • Explain a function in your codebase
  • Generate unit tests for one module
  • Summarize a long document
  • Step 2: Try a cloud agent

    Use a cloud-hosted agent so you can focus on the workflow rather than hardware:

  • [Claude Code](/agentmarketplace/claude-code) for terminal-driven coding
  • [GitHub Copilot](/agentmarketplace/github-copilot) for IDE autocomplete
  • [Cursor](/agentmarketplace/cursor) for editor-based agentic coding
  • Step 3: Define done and guardrails

    Before prompting, write down what success looks like and what the agent must not touch.

    Step 4: Compare before committing

    Use the [Agent Directory](/agentmarketplace/agents) comparison tool to evaluate 2–3 tools for your stack.

    Step 5: Go local if you need privacy

    When data cannot leave your machine, move to local models and open-source agents. See:

  • [Local Model Agents](/agentmarketplace/use-cases/local-models)
  • [Self-Hosting](/agentmarketplace/self-hosting)
  • [Deployment Recipes](/agentmarketplace/deployment-recipes)
  • Step 6: Stay current

    Follow the [AI News](/agentmarketplace/news) page and check the [Agent Changelog](/agentmarketplace/changelog) for updates.