QA Web Agent

Every feature, grouped by what it does for you.

Five groups, sixteen features. Each page shows the real inputs, the real outputs, one Jade example, and the section every honest product page needs: where it stops today.

Test

Plain-language tests, exploration, and the environments they run in.

Dashboard screenshot for the Test group

  1. Regression tests

    A test is a markdown file. An agent runs it in a real browser, finds its own prerequisites in the catalog, retries once, and returns a verdict with evidence.

    Tests · Jobs · Run groups

  2. Exploratory tests

    Every page the agent has visited can be swept against a library of named checks, one category at a time, and any failed check can become a regression test.

    Explore

  3. Site discovery

    A crawler, static analysis of the source and an agent build the page tree of your app, with grouping rules that fold ids and tenant slugs onto one page.

    Explore

  4. Isolated environments

    Each run boots your app from your repository at the exact commit, with patches, seed data and a dependency cache, in a private namespace removed when it ends.

    Environments · Projects

  5. Email flows

    A captured mailbox hands the agent the app's outbound mail as tools, so magic-link, one-time-code and invite flows are tested end to end, not left at the inbox.

    Projects

Verify

Verdicts where the work happens: pull requests, issues, evidence, trends.

Dashboard screenshot for the Verify group

  1. Pull requests

    Watch a pull request: the agent plans a suite from the diff, runs it on every push against the exact commit, and posts one label and one updating comment.

    Pull requests · Tracking

  2. Issues and filed bugs

    Watch an issue: the agent reproduces it, writes a repro test and retests on every referencing commit. A failed run becomes a filed issue with steps and video.

    Issues · Tracking

  3. Evidence

    Labelled screenshots, whole-session video, a captioned walkthrough and the full agent transcript on every run, behind links that stay valid for months.

    Jobs · Video editing

  4. Reporting and health

    What regressed and what was fixed since the last run, a morning screen for every suite, quarantine for flip-flopping tests and the dollar cost of every job.

    Suite health · Reporting · Cost

Fix

From a confirmed finding to a reviewed change.

Dashboard screenshot for the Fix group

  1. Agent code changes

    Assign a confirmed bug to the code agent. It edits, boots your app from its branch, proves the fix with the real test and hands a reviewer the diff to graduate.

    Code changes · Issues

  2. Chat with the agent

    Ask for an exploration, a reproduction, a new test, a suite run or a filed issue in conversation. The agent stands up an environment when it needs one.

    Chat

Document

Requirements and release notes that follow the code.

Dashboard screenshot for the Document group

  1. Requirements (BRDs)

    Requirements written from the codebase and your specs, updated by every merged PR, contradictions kept as conflicts, each claim scored against the code.

    BRDs · Pipeline · Conflicts · Confidence

  2. Release notes

    A daily changelog draft in plain language with categories, PR citations and screenshots from passing runs, behind a publish gate a person controls.

    Changelog

Run

Schedules, webhooks, models, integrations, many products on one platform.

Dashboard screenshot for the Run group

  1. Triggers and webhooks

    Fire a set of tests on a schedule, on a signed inbound webhook or by hand; deliver results to Slack or any endpoint as a signed payload with screenshots.

    Triggers · Webhooks

  2. Models, MCP and API

    Pick OpenAI, Anthropic, Hermes or Claude Code per project, forward your app's MCP tools to the agent, drive the platform as an MCP server, mint scoped API keys.

    Model · Settings

  3. Local worker nodes

    Add a developer machine to the fleet: the cloud's test-worker image in Docker, leasing jobs over HTTPS, with sized pools and a terminal status screen.

    Workers · Projects

  4. Many products, one platform

    Plug your product into a project of its own: your repo, environment and model. Customer feedback becomes a reproduced bug, a validated change, then a PR.

    Projects · Vendor