useglimt

CLI for Glimt — run AI-powered UX tests from your terminal or CI pipeline.

npm install -g useglimt

Setup

Generate an API key at glimt → Settings → API Keys, then configure the CLI:

useglimt config set-key glmt_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Commands

run

Create a test run and optionally wait for it to finish.

useglimt run --url https://yoursite.com --goal "Find pricing and sign up"

| Flag | Description | | ---------------------- | ---------------------------------------------------------------------------- | | --url <url> | URL to test (required) | | --goal <goal> | What the simulated users are trying to accomplish (required) | | --personas <presets> | Comma-separated persona presets, default: alex,margaret,carlos | | --watch | Stream progress until the run completes | | --json | Machine-readable JSON output (implies --watch) | | --local | Run Playwright on this machine — works with localhost and VPN-protected URLs | | --ui | Open a visible browser window during a --local run | | --browser <type> | Browser to use for --local runs: chrome (default) or firefox |

Available personas: alex, margaret, carlos, david, zoe

# Run with specific personas and wait for results
useglimt run --url https://yoursite.com --goal "Sign up for free trial" \
  --personas alex,zoe,margaret --watch

# Run against localhost
useglimt run --url http://localhost:3000 --goal "Find the pricing page" --local

# Watch it run in a visible browser window
useglimt run --url http://localhost:3000 --goal "Complete checkout" --local --ui

# Use Firefox instead of Chrome
useglimt run --url http://localhost:3000 --goal "Find pricing" --local --browser firefox --ui

watch

Stream live progress of an already-created run.

useglimt watch <runId>
useglimt watch <runId> --json

status

Check the current status of a run.

useglimt status <runId>

report

Display the full report for a completed run.

useglimt report <runId>
useglimt report <runId> --json

list

List your recent test runs.

useglimt list
useglimt list --json

cancel

Cancel a queued or running test.

useglimt cancel <runId>

config

Manage CLI configuration. Settings are stored in ~/.useglimt/config.json.

useglimt config set-key <api-key>  # Set your Glimt API key
useglimt config set-url <url>      # Override the API base URL
useglimt config show               # Print current config

CI usage

# .github/workflows/ux-test.yml
- name: Run UX test
  env:
    GLIMT_API_KEY: ${{ secrets.GLIMT_API_KEY }}
  run: |
    npx useglimt run \
      --url "${{ vars.SITE_URL }}" \
      --goal "Find pricing and sign up" \
      --personas alex,margaret,zoe \
      --watch

Local mode

With --local, Playwright runs on your machine so you can test pages that aren't publicly accessible (localhost, staging behind VPN, etc.). AI inference is proxied through the Glimt API using the platform's Anthropic key — no separate Anthropic account needed.

Requirements:

• A Glimt API key (useglimt config set-key ...)
• Browser installed via Playwright:

  npx playwright install chromium   # for Chrome (default)
  npx playwright install firefox    # for Firefox

# Headless (default)
useglimt run --url http://localhost:3000 --goal "Complete checkout" --local

# Watch it happen in a real browser window
useglimt run --url http://localhost:3000 --goal "Complete checkout" --local --ui

# Use Firefox
useglimt run --url http://localhost:3000 --goal "Complete checkout" --local --browser firefox

JSON output

Every command accepts --json for scripting. run --json emits newline-delimited events as the run progresses:

{"event":"session_start","persona":"alex","sessionIndex":0,"totalSessions":3}
{"event":"step","persona":"alex","step":0,"thinkAloud":"Looking for a sign-up button…"}
{"event":"session_complete","persona":"alex","goalReached":true,"stepCount":4}
{"event":"completed","runId":"clxxx"}

License

MIT