LEX AI commands
Reference for the LEX AI command line: lex setup-with-ai, ai-dashboard, ai-verify, ai-update, ai-faq and ai-issue-report, with their flags and what each one actually touches.
LEX AI commands
Alongside the workflow modes, LEX AI ships a small set of command-line helpers for setup, control, diagnostics, and support. Run them from the root of the LEX project you want them to act on. Every name is hyphenated: lex setup_with_ai is the Python module path, not a command, and fails with No such command.
Setup and control
Command | What it does |
|---|---|
| Wire the |
| Open the local browser control surface: pick the workflow mode, select coding environments, manage credentials, read system information. |
lex setup-with-ai accepts -p/--project-root, --github-token, --remote-mcp-api-key, --remote-mcp-url, --mcp-mode, -e/--environment (repeatable, one per environment), --list-environments, and --no-browser. lex ai-dashboard takes -p/--project-root only. See Setup and dashboard.
Diagnostics and maintenance
Command | What it does |
|---|---|
| Make the on-disk assets match what the installed packages ship — restore missing or drifted files, prune stale mode assets that would otherwise leave a mixed tool surface. |
| Upgrade the package, then run the re-runnable migration ladder against the project. See Updates. |
| Prepare a second checkout so another LEX AI chat can work on this project in parallel. Prints the folder to open the new chat in. See Handoff. |
lex ai-verify is also the MCP pre-flight: the server runs it as a subprocess with --silent at the start of every tool call. That is why --silent prints nothing on success — anything on stdout would corrupt the JSON-RPC stream the IDE speaks over. Write failures are reported even so, because a failed write is the one result you cannot afford to miss. Other flags: -p, --mode, -e/--environment, --quiet, --align-mcp-mode/--no-align-mcp-mode (off under --silent), and --strict, which turns a soft failure into a non-zero exit.
Help and support
Command | What it does |
|---|---|
| Render the FAQ locally from the installed package and open it at the Prompt Builder. |
| Build a support bundle under |
lex ai-faq takes no options and fetches nothing from the internet. The page is written from the wheel you have installed, so it always describes your version rather than a newer one; set LEX_AI_FAQ_URL only if you deliberately want a hosted copy. See FAQ.
Typical order of use
lex setup-with-ai— connect the project (once).lex ai-dashboard— confirm state, select environments, pick a mode.- Work through your chosen mode with the assistant.
lex ai-verify— if the assistant reports trouble.lex ai-update— to move to a newer build.lex ai-issue-report— if you need to escalate to support.
About the issue report bundle
The bundle collects the project .env, the Copilot MCP config, the Copilot state database and root, and VS Code workspace-storage and .vscode directories, so support sees real state rather than a description of it. Credential values are masked on the way in, before anything is written: manifest.json carries a redaction block naming each touched file and how many values were masked, and the CLI prints the total. A file too large to scan in memory is dropped rather than shipped unscanned. Without --yes the command warns and asks for confirmation; other flags are -p, -o/--output, and --artifact-mode auto|off|strict. The ZIP is written first, and submitting a ticket through the remote MCP is a best-effort extra that never blocks the bundle.
Was this helpful?
Your feedback shapes what we write next.
