Cross-product workflows

Molar's four products chain into continuous QA loops. Scenarios and run history stay in your account as work moves from discovery → CI → production → debug → fix.

Workflow 1 — Discover → test → gate PRs

The most common onboarding path:

Steps

1. Connect your app at app.molar.it (URL + optional GitHub repo).
2. Cartographer crawls and proposes .molar.md scenarios from discovered flows.
3. Approve scenarios you want in CI and production.
4. Install GitHub App — Settings → Organization → GitHub.
5. On every PR, Guard runs affected scenarios as a required status check.
6. On failure, open Trace from the check details or dashboard Runs.
7. Optionally enable Mender (mender.mode: suggestive in frontmatter) to draft a fix PR.

CLI equivalent (MCP — not CLI):

# Agent / MCP path
molar_generate_scenarios_from_url → molar scenarios validate && molar scenarios compile → molar_run_scenarios

Human CLI path after scenarios exist:

molar scenarios validate && molar scenarios compile
molar run

Or from an agent: MCP molar_generate_scenarios_from_url → molar_run_scenarios.

Workflow 2 — Hermetic checkout with Clones

Test Stripe + email without real charges or inboxes:

1. Add clone declarations to scenario frontmatter:

clones:
  - { kind: stripe, seed: stripe-base }
  - { kind: email }

2. Setup steps reset clone state and drain outbox.
3. Steps use clone webhooks (Wait for webhook stripe.checkout.session.completed).
4. Assertions query clone state (Assert Stripe Clone has 1 active subscription).

Local: molar clone stripe --start before molar run.

Cloud: MCP molar_clone_spawn or Guard worker auto-starts clones per scenario config.

Workflow 3 — Production guard + shadow diff

Keep production healthy after deploy:

1. Add schedule to scenario frontmatter (cron + regions).
2. Guard runs on interval from edge workers.
3. Failures create incidents in dashboard Inbox and Production guard.
4. Optional shadow_prod: true compares canary vs production responses without mutating prod.

Investigate via MCP:

• molar_get_incident
• molar_get_guard_status
• Prompt: investigate_incident

Workflow 4 — Debug → fix → verify

When a run fails:

1. Open Trace from run detail (app.molar.it/r/{shortId} for shares).
2. Replay timeline — video, DOM, network, console, agent ribbon.
3. Ask Debugger AI (MCP molar_debug_run or Trace UI) — bills debugger meter.
4. Mender analyzes failure + repo context → opens fix PR.
5. Review via molar mender review <id> or dashboard Auto-fix.
6. Merge fix; Guard re-runs on PR to verify.

Semantic diff between runs: molar trace diff <run-a> <run-b> or MCP molar_diff_runs.

Workflow 5 — Agent-driven QA (MCP)

For Cursor / Claude Code users:

1. Install MCP server (https://mcp.molar.cloud/mcp).
2. Agent calls molar_generate_scenarios_from_url on a feature branch URL.
3. molar_validate_scenario → molar_compile_scenario → molar_run_scenarios.
4. On failure, molar_replay_trace + molar_debug_run.
5. Prompt fix_failing_tests with PR number.

Local/air-gapped: molar mcp serve with local Clones; add --cloud-access for cloud-only tools.

Workflow 6 — OTP / auth flows

Sign-in scenarios with real OTP delivery:

1. Provision a synthetic phone or mailbox via agent comms.
2. Use Clerk clone or real auth with synthetic user (synthetic_user.pattern).
3. Cartographer OTP service polls agent comms /otp/poll.
4. Scenario fills OTP field when code arrives.

See Agent comms.

What carries between steps

Account type impact

Related

• Platform overview
• Your first scenario
• Guard · Traces · Clones product docs
• MCP server setup