Choosing Your Shiplight AI Starting Point: MCP Server, Shiplight Cloud, or the AI SDK?

AI is changing how software gets built. Teams are shipping more UI changes, more frequently, often with AI coding agents writing and iterating on code at a pace that traditional QA workflows were never designed to match.

The hard part is not generating code. It is verification.

Shiplight AI is built around a simple idea: verification should happen inside the development loop, in a real browser, while changes are being made, and the result of that verification should become durable regression coverage with near-zero ongoing maintenance.

If you are evaluating Shiplight, there are three common ways teams start:

1. Shiplight MCP Server for AI coding agents and autonomous verification
2. Shiplight Cloud for team-wide test creation, execution, management, and reporting
3. Shiplight AI SDK for teams with existing Playwright suites that want AI-native execution and stabilization

This guide helps you pick the right entry point and shows how they fit together as your quality system matures.

The three entry points, explained in plain terms

1. Start with Shiplight MCP Server if you build with AI agents

Shiplight MCP Server is designed to work alongside AI coding agents, so your agent can open a real browser, validate behavior, and close the feedback loop by turning what it verified into end-to-end tests.

Where MCP Server shines:

• UI verification during development, not after merge
• Autonomous test generation from what the agent already built and checked
• Tight feedback loops, where diagnostic context goes back to the agent for remediation

A practical signal that MCP Server is the right start: your team is already using tools like Claude Code, Cursor, or Windsurf and wants verification to happen as naturally as “run the app and check it.”

2. Start with Shiplight Cloud if you need shared visibility and operational control

Shiplight Cloud is the team platform: test creation (including no-code and natural language), suite management, scheduled runs, CI triggers, cloud execution, debugging, and reporting.

Where Shiplight Cloud shines:

• Collaborative test authoring with an editor designed for iteration
• Organized suites and repeatable runs that teams can depend on
• Actionable failure understanding, including AI-generated summaries of failed results
• CI/CD integration, including GitHub Actions workflows driven by Shiplight API tokens and suite/environment IDs

If your main pain is not “how do I create one test,” but “how do I run the right tests, at the right time, with the right visibility,” Shiplight Cloud is usually the best first move.

3. Start with Shiplight AI SDK if Playwright is already your testing backbone

If your organization already treats E2E testing as code, the Shiplight AI SDK is designed to extend, not replace, your existing Playwright approach. Tests stay in your repo, follow your normal workflows, and Shiplight adds AI-native execution, stabilization, and structured feedback.

Where the AI SDK shines:

• Minimal disruption to an established Playwright project
• AI-native actions and assertions while keeping tests deterministic and rooted in code
• Self-healing and stabilization surfaced as code changes when needed

If you have hundreds of Playwright tests and your top problem is flakiness, brittleness, and maintenance cost, the SDK path typically delivers the fastest ROI with the least workflow change.

A decision checklist you can actually use

Choose MCP Server if:

• You want verification to happen while an AI agent is building the feature
• You care about “browser proof” before review
• You want testing to scale with AI-driven development velocity

Choose Shiplight Cloud if:

• Multiple roles contribute to quality and you need a shared system of record
• You want scheduling, dashboards, and consistent execution environments
• You need AI summaries to accelerate triage and reduce time-to-diagnosis

Choose Shiplight AI SDK if:

• Your tests must remain code-first and repo-native
• You already use Playwright and want AI-native execution without a rewrite
• CI gating and deterministic behavior are non-negotiable

How the pieces fit together as you scale

Many teams adopt Shiplight in layers:

1. Verify locally with MCP Server, directly in the development workflow
2. Save what matters as readable tests, often as YAML test flows that live in the repo alongside the product code
3. Scale execution and visibility in Shiplight Cloud, with suites, scheduled runs, and CI triggers
4. Extend or modernize existing Playwright coverage with the AI SDK, especially for large code-first organizations

This layered approach matters because it prevents the usual E2E testing trap: a tool that works for a single team member, but collapses under the demands of CI, ownership, reporting, and change management.

What “tests as intent” looks like in practice

Shiplight’s documentation describes YAML test flows that can include locators when you want them, but can also rely on the agent to resolve elements from the action description when locators are not specified.

That distinction is important. It changes the authoring mindset from “bind to brittle selectors” to “express the user’s intent,” while still giving engineering teams a path to explicitness and control.

And because real applications are messy, Shiplight supports practical patterns like runtime conditionals and loops in test flows, which helps teams build coverage that reflects real user journeys rather than happy-path scripts.

Do not ignore the hard flows: auth and email

End-to-end coverage breaks down fastest in the workflows that matter most: authentication, account verification, password resets, and multi-step onboarding.

Shiplight includes an Email Content Extraction capability designed for tests that need to read incoming emails and extract content like verification codes or activation links using an LLM-based extractor, without regex-heavy plumbing.

That kind of capability is not a nice-to-have. It is often the difference between “we have E2E tests” and “we can actually trust our E2E tests on release day.”

Enterprise readiness: security, reliability, and integration

Shiplight positions itself as enterprise-ready, including SOC 2 Type II certification, encryption in transit and at rest, role-based access control, and immutable audit logs. It also publishes a 99.99% uptime SLA and highlights integrations across CI/CD and collaboration tooling.

For teams operating in regulated environments or with strict operational requirements, those foundations are often prerequisites to adopting any verification platform at scale.

The bottom line

If you want the simplest way to start, start where verification is currently weakest:

• If changes are built by AI agents and regressions slip in before review, start with Shiplight MCP Server.
• If you need a shared system for suites, schedules, CI runs, and triage, start with Shiplight Cloud.
• If you already run Playwright at scale and want AI-native stability without a rewrite, start with the Shiplight AI SDK.

Shiplight’s promise is not “more tests.” It is faster verification, better regression coverage, and less maintenance drag, so teams can ship faster without breaking what users rely on.