InspectorX for VS Code

🕵️ Deep, multi-agent code review inside VS Code. Less guesswork. More evidence.

InspectorX turns GitHub Copilot Chat into a structured review system that can inspect code paths, track execution live, write artifacts, and push outcomes back to Azure DevOps.

✨ Why It Feels Different

🚀 Highlights

• 💬 @inspectorx chat participant with /review and /publish
• 🌊 Flowing dependency-graph execution instead of flat prompt chaining
• 📊 Sidebar sessions, progress panel, artifacts, diagnostics, and verdicts in one workflow
• 🔒 Isolated review worktrees for safer runs
• 🔗 Optional Azure DevOps publishing and feedback submission

🖼️ Dependency Graph Preview

⚡ Start In 60 Seconds

1. Install InspectorX and make sure GitHub Copilot Chat is installed.
2. Open the repository you want to review in VS Code.
3. Start a review from the sidebar or with @inspectorx /review.
4. Watch progress, inspect findings, then publish or keep local.

🎯 What You Get

🧭 Review Flow

• repo selection and session tracking in the InspectorX sidebar
• dedicated progress panel with graph and step-level status
• verdicts, findings, and diagnostics surfaced directly in VS Code

🧾 Evidence Trail

• local session artifacts under ~/InspectorX/artifacts/
• per-agent reports, progress snapshots, logs, and raw outputs
• easier debugging when a review needs explanation, not just a verdict

🤝 Team Workflow

• Azure DevOps publishing for review output
• built-in bug reporting and feature suggestion commands
• structured output that is easier to share and act on

📦 Artifact Layout

~/InspectorX/artifacts/{repo}/chat-participant/{session}/
  verdict.md
  progress.json
  session.log
  raw_results.json
  Agents/
    Profiler.md
    Analyst_Patterns.md
    Judge.md
    ...

🔧 Azure DevOps Setup

InspectorX attempts to configure ADO MCP automatically on activation by updating mcp.servers and enabling chat.mcp.autoStart. When ADO is available, it can:

• read PR and work-item context
• publish findings back to ADO
• support suggestion-style review flows
• submit feedback work items

If auto-configuration does not appear, add the same server manually in either:

• workspace .vscode/mcp.json, or
• user settings.json under mcp.servers

Replace <your-ado-org> with your actual ADO organization (for example microsoft, msazure). Use the current official package @azure-devops/mcp.

Example .vscode/mcp.json:

{
  "servers": {
    "ado": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@azure-devops/mcp",
        "<your-ado-org>",
        "-d",
        "core",
        "repositories",
        "wiki",
        "work-items"
      ]
    }
  }
}

Keep the server key as ado.

InspectorX expects the MCP server key ado. If you previously used keys such as ado-mcp or azure-devops, rename/remove them or rerun setup so the tool namespace stays on the supported mcp_ado_* path.

💡 Recommended Settings

InspectorX invokes multiple tools during a review session. To avoid repeated approval dialogs, add this to your VS Code user settings:

{
  "chat.tools.autoApprove": true
}

This allows all tool invocations to proceed without manual confirmation, similar to /allow-all in CLI-based workflows.

🛠️ Commands You Will Actually Use

• InspectorX: Start Review
• InspectorX: Open Review Progress
• InspectorX: Publish Findings to ADO
• InspectorX: Keep Findings Local Only
• InspectorX: Open Artifacts Folder
• InspectorX: Report Bug
• InspectorX: Suggest Feature

📚 Read Next

• docs/README.md: documentation index
• docs/getting-started/user-quickstart.md: step-by-step first-run setup
• docs/reference/features.md: core capabilities and UI surfaces
• docs/reference/dependencies.md: required and optional dependencies
• CHANGELOG.md: release history

🩺 Troubleshooting

• ❌ Missing Copilot Chat: install github.copilot-chat
• ⚙️ Setup looks incomplete: reload VS Code (setup runs automatically)
• 🔗 ADO features unavailable: review docs/reference/dependencies.md and docs/getting-started/user-quickstart.md
• 🐞 Found an InspectorX bug or missing workflow: use InspectorX: Report Bug or InspectorX: Suggest Feature
• 📜 Need logs: open View -> Output -> InspectorX

InspectorX is built for review workflows that need traceability, not just suggestions.