1Code Documentation

Use these patterns to resolve common re-render issues identified by WDYR:

Pattern 1: Deeply equal objects/arrays with different references

Problem: diffType: "deepEquals" indicates objects are identical in value but new references are created every render. Fix: Wrap creation in useMemo.

Pattern 2: Inline Objects in Render

Problem: Creating objects directly in the component body. Fix: Memoize the object using useMemo.

Pattern 3: Callback Reference Changes

Problem: Defining arrow functions inline in props. Fix: Use useCallback to maintain a stable function reference.

Pattern 4: useEffect with Object Dependencies

Problem: Passing an inline object to a useEffect dependency array. Fix: Memoize the object or use primitive values as dependencies.

1Code provides full Model Context Protocol (MCP) lifecycle management through the UI.

Capabilities:

• Server Management: Toggle, configure, and delete MCP servers directly from the interface.
• Plugin Marketplace: Browse and install plugins with one click.
• Tool Interaction: View MCP tool calls with formatted inputs and outputs.
• Direct Referencing: Use @ mentions in the chat input to reference MCP servers.

To start a development environment, install dependencies, download the required agent binaries, and run the dev command.

bun install
bun run claude:download  # First time only
bun run codex:download   # First time only
bun run dev

Automations allow you to trigger agents from external events. Note that automations require a Pro or Max subscription.

Supported Triggers:

• @1code Triggers: Tag @1code in GitHub, Linear, or Slack to start an agent.
• Git Event Triggers: Run automations on push, PR, or any other git event.

Configuration Options:

• Conditions & Filters: Control the specific criteria required for an automation to fire.
• Silent Mode: Toggle whether the agent responds to the trigger for background automations.

To build 1Code for free, you must install the required agent binaries (Claude and Codex) during the build process.

Prerequisites:

• Bun
• Python 3.11 (recommended for native module rebuilds)
• setuptools (required if using Python 3.12+ via pip install setuptools)
• Xcode Command Line Tools (macOS users)

Build Steps:

1. Install dependencies.
2. Download the required agent binaries.
3. Build and package the application for your target platform.

# Prerequisites: Bun, Python 3.11, setuptools, Xcode Command Line Tools (macOS)
bun install
bun run claude:download  # Download Claude binary (required!)
bun run codex:download   # Download Codex binary (required!)
bun run build
bun run package:mac  # or package:win, package:linux

To debug infinite re-render loops and unnecessary re-renders in the desktop app using the Why Did You Render (WDYR) library, follow these steps:

1. Open src/renderer/wdyr.ts.
2. Set const WDYR_ENABLED = true.
3. Run bun run dev.
4. Reproduce the issue; re-render logs will appear in the console.

For WDYR to track all components (not just those wrapped in React.memo or PureComponent), you must configure it as the JSX import source in electron.vite.config.ts during development mode.

This is a critical configuration step for full coverage.

react({
  jsxImportSource: isDev
    ? "@welldone-software/why-did-you-render"
    : undefined,
})

You can launch 1code directly from your terminal by passing a directory path as an argument. This allows you to open specific projects immediately.

Usage:

• Open the current directory: 1code .
• Open a specific project: 1code /path/to/project

WDYR Not Logging

1. Verify WDYR_ENABLED = true in src/renderer/wdyr.ts.
2. Ensure you restarted the dev server after changes.
3. Confirm jsxImportSource is correctly set in electron.vite.config.ts.
4. Ensure wdyr.ts is the first import in src/renderer/main.tsx.

Too Many Logs

Adjust the threshold and window in src/renderer/wdyr.ts:

const THRESHOLD = 10  // Increase to reduce noise
const TIME_WINDOW = 1000  // Time window in ms

Crash Before Logs Appear

If the app crashes before WDYR can log, try:

1. Adding a debugger statement at the top of the suspected component.
2. Using the React DevTools Profiler to identify the looping component.

The application integrates with the @anthropic-ai/claude-code SDK using dynamic imports. It supports two distinct operational modes:

1. Plan Mode: A read-only mode for planning and analysis.
2. Agent Mode: A mode with full permissions for executing tools like bash commands and file edits.