Shiplight AI

Visit Website

Introduction

Shiplight AI describes itself as “the agent-native testing platform”. According to the official site, it builds and maintains your end-to-end suite: agents walk your app in a real browser and write tests in plain language you can read and change.

Shiplight installs into your coding agent as an MCP server plus a set of skills, with documented setup for Claude Code, Cursor, Codex, VS Code and Antigravity. The browser MCP lets a coding agent verify a change, for example opening a checkout page, checking the console and capturing screenshots.

Features

1. MCP plus skills: Install with npx commands into supported coding agents.
2. Real-browser verification: Agents open your app, run the flow and report console state and screenshots.
3. Plain-language tests: Tests are readable and editable by people.
4. Self-healing: Keeps suites green as the UI changes.
5. Accessibility review: Ask the agent to review a page for accessibility issues.
6. Bring your own LLM: Use your own model key and run on your own infrastructure, even on the free plan.
7. Enterprise controls: SOC 2, audit logs, private cloud and VPC deployments, dedicated CSM.

Pros

1. Every feature on free: The site says every core feature is on the free plan.
2. Agent-native: Fits teams where coding agents do most of the work.
3. Readable tests: Plain-language tests are easy to review.
4. Self-hosting option: Run on your own infrastructure.
5. Pay-as-you-go on Pro: Runs do not stop when the balance hits zero.

Cons

1. Credit model: Usage draws from credits; estimate before scaling.
2. Web focus: Browser testing is the core.
3. Young vendor: Fewer public case studies.
4. Agent-dependent setup: Best results assume a coding agent workflow.
5. Plain-language ambiguity: Natural-language steps can be interpreted differently over time.

Usage and Links (Example)

From the official site, for Claude Code:

npx -y skills add ShiplightAI/agent-skills-v2 -a claude-code -y
npx -y add-mcp "npx -y @shiplightai/mcp@latest" -n shiplight --env PWDEBUG=console -a claude-code -y

Then ask the agent: “Verify my change to the checkout flow.” It opens the page in a real browser and returns results with screenshots.

Resources:

• Shiplight website
• Pricing

Pricing

From Shiplight’s pricing page:

1. Free: $0, free forever. Every core feature, no credit card, a one-time $10 signup bonus, bring your own LLM key and infrastructure.
2. Pro: $60 a month billed monthly (yearly saves 17%). Monthly credit plus pay-as-you-go.
3. Enterprise: Contract priced. Volume pricing, SSO, SOC 2, dedicated CSM.

Recommended for?

1. Teams building with coding agents that want verification inside the agent loop.
2. Startups wanting a free plan with every core feature.
3. Engineers who want readable tests that people can edit.
4. Not a fit for native mobile testing.

Conclusion

Shiplight AI is built for agentic development, with MCP and skills, real-browser verification and a generous free plan.

Keep an eye on credits and check readable tests stay precise as they heal.