Supported AI Tools

Inspecto bridges the gap between your browser and your AI assistant. It supports multiple interaction modes depending on how your AI tool is architected.

When you use the structured onboarding flow or npx @inspecto-dev/cli init, the CLI currently probes two install surfaces automatically: CLI tools and IDE extensions. Built-in IDE targets are still supported by the runtime, but they are not auto-detected by the CLI onboarding probe yet.

Onboarding

For most users, the default path is assistant-first onboarding:

1. Run npx @inspecto-dev/cli integrations install <assistant> --host-ide <ide> from the target project root.
2. Follow the onboarding flow opened by the CLI.
3. If onboarding did not open automatically, ask the assistant to set up Inspecto in this project.
4. Use npx @inspecto-dev/cli integrations doctor <assistant> --host-ide <ide> --compact only when you want to check blockers before install or troubleshoot.

Then open the app in the browser, use the launcher for Inspect mode or Annotate mode, and use Alt + Click anytime for Quick jump.

If you are setting up Inspecto manually in a terminal, use npx @inspecto-dev/cli init as the guided fallback.

If you are building your own agent/runtime integration, use the structured onboarding flow directly:

1. inspecto onboard --json
2. if status is needs_target_selection, explain that this step chooses which local development build target should receive Inspecto, then rerun with --target <candidateId> using one returned target candidate. The CLI also accepts a returned configPath as a compatibility fallback.
3. if status is needs_confirmation, confirm the planned changes and rerun with --yes
4. complete the ideExtension step if using IDE mode (auto-install when possible, otherwise show install links/commands)
5. then follow verification guidance to restart or prompt for dev-server validation
6. if status is error, run inspecto doctor --json

For field-level response semantics and status handling, see Onboarding Integrations and the onboarding command docs.

When onboarding configures agent-driven annotate delivery, the JSON result may include a structured runtime handoff:

{
  "handoff": {
    "dailyUsage": {
      "mode": "agent",
      "skill": "inspecto-agent",
      "prompt": "Use $inspecto-agent to claim Inspecto tasks continuously",
      "requiresMcp": true
    }
  }
}

If handoff.dailyUsage is present, prefer it over hard-coded follow-up wording. It is the canonical post-onboarding entrypoint for daily MCP-driven annotation work.

Interaction Modes

Use the mode names as delivery routes:

Route	Best for	Requires IDE extension?
IDE route	Immediate Inspect / Annotate prompt handoff to an AI panel	Yes, except built-in IDE targets
CLI route	Sending prompts into a terminal-based assistant	Usually yes, to open/manage the terminal
MCP route	Durable annotation sessions, browser timeline, custom workflows	No
Browser-only / Clipboard	Copying context manually or using unsupported editors	No

If you want custom workflow buttons such as deploy, PR, release, or test automation, prefer the MCP route. It lets Inspecto queue a workflow session and lets the agent execute it with its own skills, MCP servers, and tools.

1. Extension Mode

The AI tool is installed as an IDE extension (e.g., in VS Code). Inspecto will use the IDE's custom URI schemes to dispatch the prompt to the AI chat panel.

Supported:

• GitHub Copilot (copilot.extension)
• Claude Code (claude-code.extension)
• Gemini Code Assist (gemini.extension)
• CodeX (codex.extension)

2. Built-in Mode

The AI tool is natively integrated into a fork of the IDE. No extensions are required. Inspecto will open the local file and trigger the native AI chat panel.

Supported:

• Cursor (cursor.builtin)
• Trae (trae.builtin)
• CodeBuddy (codebuddy.builtin)

3. CLI Mode

The AI tool runs entirely within the terminal. Inspecto will open a new terminal panel in your IDE, launch the CLI tool (if not already running), and paste the prompt.

Supported:

• Claude Code CLI (claude-code.cli)
• Trae CLI / Coco (coco.cli)
• Gemini CLI (gemini.cli)
• CodeX CLI (codex.cli)

Switching Providers

The structured onboarding flow and init command will probe available CLI tools and VS Code-compatible extensions, then prompt you to select the default target. Workspace extension recommendations may also appear in that probe, so the result is best interpreted as available integration candidates rather than a strict list of locally installed AI tools.

To change it later, simply update your .inspecto/settings.local.json:

{
  "provider.default": "claude-code.cli"
}

Support for Standalone / Clipboard Mode

If you don't want to use IDE extensions or CLI terminals, you can configure Inspecto to use a standalone mode:

{
  "ide": "none"
}

In this mode, Inspecto relies on:

1. MCP (Model Context Protocol) to deliver annotation sessions to running agents (see MCP Integration)
2. Clipboard integration with a "Copy Context" button that writes formatted Markdown directly to your clipboard so you can paste it anywhere.