AWT (AI Watch Tester)

Install

Works with Claude Code, Codex, VS Code, Cursor, Windsurf, Claude Desktop. Read the repository before you give any tool access to your data.

npx agentmarket install connector/ksgisang-awt

About

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this server can do and what it asks for, from its tool list, install config and code. Checked Oct 25, 2026.

Tools

7 read · 2 write · 0 delete · 3 not yet classified

Secrets

2 (API keys, tokens or passwords)

Install scripts

None found

Runs as

Python package

Tools (12)

Read from the source code; the running server may differ. Descriptions are the server’s own words; hidden characters are shown as [U+…].

• aat_cost

  View AWT AI API usage costs. Shows daily/monthly cost breakdown by provider, including token counts and estimated USD costs.

• aat_devqa

  Full auto DevQA loop: scan → generate → show scenario → wait for approval → run → fix. This is the recommended all-in-one command for most testing tasks. It will: 1. Scan the target URL for interactive elements 2. Generate a test scenario from scan data 3. Show the scenario and PAUSE for human approval ([Enter] in terminal) 4. Run the test with real Chrome 5. On failure: re-scan, fix scenario, retry (up to 5 times) 6. Report final results ⚠️ After calling this tool, tell the user to switch to their terminal and press [Enter] to approve the generated scenario and start the test. There is no w…

• aat_diff

  Compare current screenshots against saved baselines (visual regression). Runs the scenario, takes screenshots, and compares each step against the baseline captured by ``aat_snapshot``. Reports SSIM similarity scores and generates diff images for failed steps. No AI tokens used — pure OpenCV image comparison. AFTER calling this tool: 1. Review the SSIM scores and diff images 2. Steps below threshold are marked FAIL 3. Diff images are saved to .aat/diffs/ Args: scenario_file: Path to a YAML scenario file or directory. url: Target URL override (optional). threshold: SSIM threshold…

• aat_doctor

  Check AWT environment and dependencies. Verifies: Python version, Playwright browsers, Tesseract OCR, AI provider connectivity. Run this first to diagnose setup issues.

• aat_generate_from_doc

  Generate test scenarios from a document (spec, plan, or any text file). Reads a document file and uses AI to generate E2E test scenarios. Supports: Markdown (.md), plain text (.txt), and any structured text including GSD PLAN.md files, PRDs, or free-form spec documents. This is useful for: - GSD workflow: pass your PLAN.md to auto-generate Verify scenarios - Spec-driven testing: upload a PRD and get test cases - Harness integration: use as a quality gate in your AI coding workflow AFTER calling this tool: 1. Read the generated scenario YAML files 2. Present them to the user for review 3. As…

• aat_list_scenarios

  List YAML scenario files in the current directory. Scans for .yaml and .yml files in scenarios/ directory and current directory. Returns file paths and scenario names.

• aat_run

  STEP 3: Run AWT test scenarios (requires prior user approval). ⚠️ IMPORTANT: Only call this AFTER: 1. You have scanned the site with aat_scan (Step 1) 2. You have shown the scenario to the user (Step 2) 3. The user has explicitly approved the test (e.g. "go ahead", "yes", "run it") If a test step fails, STOP and report to the user. Ask what to fix. Do NOT auto-retry or auto-fix without user permission. Exit code 3 with STATUS: WARNINGS means every step ran but at least one changed nothing: a click that missed its target, or a find_and_type whose field was a <select>, readonly, disabled, or …

• aat_run_skill_mode

  STEP 3 (alternative): Run AWT in skill mode with structured failure diagnosis. Same rules as aat_run — requires prior user approval. ⚠️ IMPORTANT: Only call this AFTER the user has approved the scenario. On failure, outputs a structured === AWT SKILL DEVQA === block with: SCENARIO, FAILED_STEP, ERROR, ACTUAL_CAUSE, SCREENSHOT path, POSSIBLE_CAUSE. ERROR is the step's own message — what the scenario author expected, written before the run. ACTUAL_CAUSE is what actually happened, and appears only when it differs from ERROR. Diagnose from ACTUAL_CAUSE whenever it is present; reading ERROR as …

• aat_scan

  STEP 1: Scan a web page to discover all interactive elements. This is the FIRST step of the AWT workflow. Call this BEFORE creating any test scenarios. It analyzes the page and saves results to .aat/scan_result.json. AFTER calling this tool: 1. Read the output to see what elements were found 2. Present a summary to the user 3. Ask the user: "Should I create a test scenario based on these elements?" 4. WAIT for user approval before proceeding ⛔ Do NOT proceed to aat_run without user approval. Args: url: The URL to scan (e.g. http://localhost:3000)

• aat_snapshot

  Capture baseline screenshots for visual regression testing. Runs the scenario and saves after-screenshots as baselines. Use ``aat_diff`` later to compare current state against these baselines. No AI tokens used — pure Playwright + OpenCV. Args: scenario_file: Path to a YAML scenario file or directory. url: Target URL override (optional). responsive: Capture 3 viewports (mobile/tablet/desktop). viewport: Single viewport WxH (e.g. '375x812'). Overrides responsive. console: Collect browser console errors during capture. console_fail: Fail if console errors detected. Imp…

• aat_validate

  Validate YAML scenario files against AWT schema. Checks for: missing required fields, invalid action types, malformed targets, and other schema violations. Args: scenario_file: Path to a YAML scenario file or directory.

• aat_watch

  Start watch mode — auto-run tests on file changes. Monitors source files and scenario files for changes. When a change is detected, automatically re-runs tests. ⚠️ This is a long-running command. Tell the user it will keep running until they press Ctrl+C in their terminal. Args: scenario_file: Path to scenario file or directory. url: Target URL override (optional). watch_dir: Additional directory to watch (optional).

Compare and alternatives

Featured in

• Best MCP servers for software testing

Signals

Stars

1

Forks

0

Open issues

0

Weekly downloads

138

Last commit

Oct 16, 2026 (13 days ago)

Last release

Oct 15, 2026 (14 days ago)

Last package publish

Oct 15, 2026 (14 days ago)

From GitHub (fetched Oct 24, 2026) and the package registry.

Details

Registry name

io.github.ksgisang/awt

Version

1.8.0

Auth

API key

Packages

pypi aat-devqa

License

AGPL-3.0

Repository

github.com/ksgisang/awt-skill

First seen

Oct 24, 2026

Last checked

Oct 24, 2026

Source

Official MCP Registrygithub-graphql