AI dashboard

What lex ai-dashboard shows in its four panels, what pressing Save actually changes, and a pre-flight routine to run before every assistant request.

A
Written By Atakan KayaLast updated 3 months ago

AI dashboard

lex ai-dashboard opens a local, browser-based control surface for the MCP server. It is the normal place to inspect configuration and select the active workflow mode before you start an assistant request.

lex ai-dashboard

It takes one option, -p/--project-root; run it from the project you want the assistant to work on. The command starts a small HTTP server bound to 127.0.0.1 on a random port, opens your browser, and stays up for at most an hour or until you press Ctrl+C. Nothing leaves your machine.

What is on the page

Four panels in one form. Fill in what you need and press Save changes once.

Panel

What it holds

Workflow Mode

Ten mode cards: Project Brief, New Project, Documentation, Edit, Review, Test Suite, Input Change, Deployment, MVP Generator, MVP Completion

Agentic Coding Environments

A multi-select over the seven supported coding tools, plus a read-only "MCP registration per config file" table

Authentication Tokens

GitHub token and Lex MCP access key, shown masked, and the remote MCP URL in full

System Information

Seven read-only rows: installed version, active mode, pending mode override, Python executable, project root, .env path, mcp.json path

The registration table labels every config file it knows about as registered, missing entry, config not created, not used, or overrides the project config, with that file's scope and mode. That last label is how you find a user-scope entry quietly outranking the project one.

A promo card links to the Prompt Builder, which the dashboard renders locally from the same FAQ page as lex ai-faq.

What saving actually does

  • Changing the mode — writes the one-shot mode override, syncs .env and the discoverable MCP configs, re-verifies the on-disk assets, and stops the server so your IDE restarts it on the new tool surface.
  • Deselecting a coding tool — unregisters lex-mcp-local from that tool's config and prunes the assets LEX AI generated for it. At least one tool must stay selected.
  • Entering a credential — writes it to .env and to the server's env block, then stops the server so the new value takes effect. A blank field keeps the current value.

A reliable pre-flight routine

  1. Open the dashboard and confirm the resolved project root is the project you mean.
  2. Confirm the credentials the work needs are present.
  3. Select the mode that matches your intent — see Choose the right path.
  4. Check that the tool you actually use shows as registered.
  5. Let the IDE restart the server, then start a fresh assistant request.

Why "select mode here" matters

Only one mode's tools are visible at a time, so choosing the mode here is what makes the right tools appear in your IDE. If the assistant reports a missing or stale tool, come back, confirm the mode, and let the surface refresh.

There is no separate "clear the IDE tool cache" button. GitHub Copilot in JetBrains caches the tool list beside its config, and LEX AI clears that cache as a side effect of an actual mode change — never as a standalone action. See Setup and dashboard for how the project was wired up in the first place.

Was this helpful?

Your feedback shapes what we write next.