OpenCode Guide
Navigate: Use Cases

Introduction

Getting Started

Guides & Integrations

Comparisons

OpenCode Use Cases

OpenCode is more than just a chatbot; it's a scripting engine driven by language models. Here are common patterns used by the community.

1. The Legacy Refactor

Scenario: You have a 500-line Python file that needs to be broken into modules. Why OpenCode: It understands file structure and can move contents while updating imports.

opencode start "Take main.py. Split the database logic into db.py and the utils into utils.py. Update all imports."
  • Advantage: OpenCode will run pytest after the move (if you ask) to ensure it didn't break anything.

2. Test Driven Development (TDD) Loop

Scenario: You want to implement a new feature rigorously. Workflow:

  1. Writer a failing test file feature.test.ts.
  2. Run the agent.
opencode run "Run npm test. Read the failure. Implement the code in feature.ts to make the test pass."
  • Advantage: The agent gets immediate feedback from the CLI output provided by the test runner.

3. Documentation "Gardener"

Scenario: Your README.md is out of date with the actual API. Why OpenCode: It acts as a reader that can diff code vs docs.

opencode run "Read all exported functions in src/api. Update docs/API.md to match the current signatures."

4. The "Translator" (Language Migration)

Scenario: Porting a React Class component to a Hook.

opencode run "Read Component.jsx. Rewrite it as a Functional Component using Hooks. Keep the same prop types."

5. Security Audit (Local)

Scenario: Checking for hardcoded secrets before commit. Why OpenCode: Using a local model (Ollama) means you aren't sending your potential secrets to OpenAI to be checked.

# Using local Llama 3
opencode run "Scan src/ for any strings that look like API keys. Report line numbers."

6. MCP Power User: Database Chat

With the Postgres MCP Server configured:

  • "Show me the schema for the 'users' table."
  • "Write a migration to add a 'last_login' column."
  • "Generate a seed script with 50 fake users."

The agent queries the DB to understand the types before writing code.

Next Steps