Agents & Monitoring

UnfoldAI monitors AI coding agents that operate inside VS Code. Detection is automatic — you don't need to configure anything for supported agents.

Supported Agents

Agent	Detection	Notes
Cursor	Automatic	Full terminal + file monitoring
Claude Code	Automatic	Full terminal + file monitoring, MCP integration available
GitHub Copilot	Automatic	Monitors Copilot agent mode activity
Cline	Automatic	Terminal-based agent monitoring
Continue	Automatic	Terminal-based agent monitoring
Codeium / Windsurf	Automatic	Terminal-based agent monitoring
Any terminal-based agent	Automatic	If it runs commands or changes files in VS Code, UnfoldAI can monitor it

How Detection Works

UnfoldAI identifies active agents automatically based on their activity patterns within VS Code. No API keys, tokens, or manual registration are required.

If you work with a custom or niche agent that UnfoldAI doesn't recognize by default, you can add custom fingerprints:

// settings.json
{
  "unfoldai.agentFingerprints.override": {
    "my-custom-agent": {
      // Custom agent identification configuration
    }
  }
}

Multi-Agent Monitoring

If you run multiple agents simultaneously (for example, Cursor and Claude Code in the same session), UnfoldAI tracks them in a unified timeline. The Dashboard shows:

• Which agents are active
• Activity attribution per agent
• A combined timeline of all events
• Episodes that may span multiple agents

You don't need to switch between views or configure per-agent monitoring.

What "Monitoring" Includes

UnfoldAI passively observes:

• Commands: What commands agents run, their output, exit codes, and duration
• File operations: Files created, modified, and deleted by agents
• Error sequences: When commands fail, when the same error repeats, when agents appear stuck
• Dependency changes: Lockfile and manifest modifications (package.json, requirements.txt, etc.)
• Session patterns: Overall session health, risk accumulation, and behavioral trends

All monitoring happens locally within VS Code. See Privacy & Data Handling for details on what stays local and what is sent to UnfoldAI servers.

Language & Framework Support

UnfoldAI operates at the terminal and file-system level, making it language-agnostic. It works with any programming language, framework, or toolchain that your agents use. There are no language-specific features or limitations.

Remote Development

UnfoldAI is designed for local VS Code environments. Support for remote development scenarios (SSH, WSL, GitHub Codespaces, Dev Containers) is best-effort. Terminal monitoring behavior may vary in remote contexts. If you encounter issues with remote setups, please report them.