Skip to main content

Tester Quick Start

Go from zero to a green test in about five minutes — through the web UI, or through MCP for AI‑assisted and scripted testing. Both drive the same platform and the same tests.

Before you start — one credential​

Tests authenticate using a credential stored in the environment. Without it, an authenticated test refuses to run (by design — the platform never runs a test unauthenticated and calls it green). Your admin sets this once per environment:

  • Web UI: Environment → Credentials → Add (username/password, or a static token)
  • It appears as a tag (e.g. admin), secret masked, marked default.

Supported credential types: BASIC (username/password), BEARER (token), API_KEY (name + value), OAuth (client id + secret).

Lane A — Web UI (nothing to install)​

  1. Open https://app.proofarc.ai and sign in.
  2. Projects → open a project (e.g. User Management Platform).
  3. Pick a test and Run:
    • API — a scenario like User CRUD → Run → watch 5/5 steps go green (create → get → update → delete → 404).
    • UI — a UI test → Run → watch the browser drive and the steps pass.
    • Performance — open an API scenario → Convert to Performance → add load phases → Run → see p95 / error‑rate against your SLO.
  4. Green means done. Open the run for the per‑step trace and proof links.

Lane B — MCP (AI‑assisted or scripted)​

Use this with Claude Desktop, any MCP client, or the Python client.

1. Get a token

export TOKEN=$(curl -s https://app.proofarc.ai/api/auth/login \
  -H 'Content-Type: application/json' \
  -d '{"username":"<your-user>","password":"<your-password>"}' | jq -r .token)

2. Connect your MCP client to:

https://app.proofarc.ai/mcp      (Authorization: Bearer $TOKEN)

3. Sanity check — list_projects returns your projects (e.g. Default Project, User Management Platform).

4. Run your first test

find_scenarios(project="User Management Platform")     → pick a scenario id
execute_scenario(scenario=<id>, environment=<env>, credential_tag="admin",
                 wait=true, summarize=true)

Expect status: COMPLETED with every step passed.

5. One tool per test type

TypeToolsGreen looks like
APIexecute_scenario5/5 passed (create → get → update → delete → 404)
Performanceconvert_api_to_performance_scenario → run_performance_scenario → get_performance_resultverdict: PASS, p95 under threshold
UIcrawl_by_target → create_ui_test_from_yaml → run_ui_testpassed > 0, failed: 0

Full scripted example: Python client walkthrough.

What "green" guarantees​

The platform is built so a green result is a real green:

  • Auth is required, not optional. A test that needs a credential and can't resolve one refuses instead of running unauthenticated. Fix: add the credential to the environment.
  • No false greens. A 100%‑error run never reports PASS. A performance run that breaches its SLO reports FAIL; a 100%‑error run with no SLO reports NO_SLO_DECLARED, passed: false — never a silent pass.
  • Performance is the full flow. Converting an API scenario carries the whole CRUD lifecycle under load — writes included, created ids chained through the flow, and unique data per virtual user so concurrent creates don't collide. Not just a GET.

Next steps​