Help / Blue Codens

Blue Codens setup

QA automation. Generates Playwright E2E tests from natural-language requirements, runs them, and returns results (video / trace / network log). Failures can chain into Red Codens for an auto-fix.

Overview

Blue Codens' core unit is an E2E test. Generated from natural-language requirements and executed in Blue's sandboxed browser runners.

Generation: Claude API + Playwright codegen, with bias toward stable selectors.
Execution: Blue's Playwright runner with parallel execution (concurrency varies by plan tier).
Results: pass / fail status, video, trace, network log, console log.
Failure chaining: hand off failures as Red bug reports for an auto-fix PR via Red → Purple.

Prerequisites

A Codens account and organization.
codens-mcp installed and logged in.
The target URL must be reachable from Blue's runners. Add Blue's outbound IPs to your allow list if needed.
Optional: test-user credentials, stored in Blue's encrypted secret store.

Step 1. Create a project

From the dashboard (app.blue.codens.ai) → "New Project", or via MCP together with other products:

await tools.codens_register_project_unified({
  repo_full_name: "Corevice/your-repo",
  products: ["blue"]
});

A project can hold multiple base URLs (prod / staging / dev) and switch between them per run.

Step 2. Register a test environment

Dashboard → Project → Environments → "Add environment".
Set a base URL (e.g. https://staging.your-app.example.com).
Optionally save test-user credentials as secrets — injected as environment variables at run time.
If your service has an IP allow list, copy Blue's outbound IPs (shown in the dashboard) into it.

Be careful with destructive tests on production. Mutating flows like "sign up" should run against staging or a dedicated environment that accepts the X-E2E-Run-ID header for filtering.

Step 3. Create your first E2E test

// Generate from a natural-language requirement
const t = await tools.blue_e2e_test({
  action: "generate",
  project_id: "...",
  requirement: "Verify that a user can complete a checkout flow with an annual plan.",
  language: "playwright"
});

// Execute
await tools.blue_e2e_test({ action: "run", e2e_test_id: t.id });

// Fetch results (pass/fail, video URL, trace URL, network log)
const result = await tools.blue_e2e_test({
  action: "get_results",
  e2e_test_id: t.id
});

The generated Playwright script is editable from the dashboard. Subsequent runs only need action: "run".

Operational notes

Red chain: turn failed tests into bug reports and push them through Red's analysis + fix PR loop.
Run-ID injection: Blue automatically attaches an X-E2E-Run-ID header for traceability, useful for filtering mutating tests in your backend.
Webhook (Blue → Purple): a Purple workflow can also call into Blue as its verify step.
Trace / video retention: depends on plan tier; longer retention starts at Pro.

Next steps

Blue use case

Generate / run / inspect E2E

Red chain

Failed tests → fix PR

Blue tools (1)

Full action list