Usage

September 8, 2026 ยท View on GitHub

How It Works

AgentsBridge uses a file-based protocol for communication:

  1. Command File: AI agents write commands to .agents-unity-bridge/command.json
  2. Processing: Unity Editor polls for commands and executes them
  3. Response File: Results are written to .agents-unity-bridge/response-{id}.json

This approach enables multi-project support - each Unity project has its own .agents-unity-bridge/ directory, allowing multiple agents to work on different projects simultaneously.

Commands

Run Tests

{
  "id": "test-001",
  "action": "run-tests",
  "params": {
    "testMode": "EditMode",
    "filter": "MyProject.Tests"
  }
}

Parameters:

  • testMode - "EditMode" or "PlayMode"
  • filter (optional) - Test filter pattern

Compile Scripts

{
  "id": "compile-001",
  "action": "compile",
  "params": {}
}

Refresh Assets

{
  "id": "refresh-001",
  "action": "refresh",
  "params": {}
}

Get Editor Status

{
  "id": "status-001",
  "action": "get-status",
  "params": {}
}

Get Console Logs

{
  "id": "logs-001",
  "action": "get-console-logs",
  "params": {
    "limit": 20,
    "filter": "Error"
  }
}

Parameters:

  • limit (optional) - Maximum number of logs to retrieve (default: 50)
  • filter (optional) - Filter by log type: "Log", "Warning", or "Error"

Response Format

All commands return a response in this format:

{
  "id": "test-001",
  "status": "success",
  "action": "run-tests",
  "duration_ms": 1250,
  "result": {
    "passed": 410,
    "failed": 0,
    "skipped": 0
  }
}

Status Values:

  • running - Command in progress
  • success - Command completed successfully
  • failure - Command completed with failures
  • error - Command execution error

Tools Menu

The package adds menu items under Tools > Unity Bridge:

  • Show Status - Display current bridge status in console
  • Cleanup Old Responses - Delete response files older than 1 hour
  • Reset Processing State - Clear stuck processing state

Using with AI agents

Once the skill is installed, simply ask your AI agent naturally:

  • "Run the Unity tests in EditMode"
  • "Check if there are any compilation errors"
  • "Show me the last 10 error logs from Unity"
  • "Refresh the Unity asset database"

See skill/SKILL.md for complete documentation.

Complete Command Reference

For detailed command specifications including all parameters, response formats, and error scenarios, see skill/references/COMMANDS.md.