Cycling through all six. Tap any point to stop.

Measured on six things

11 deterministic insurance MCP tools; customer AI/OAuth. CA/TX only.

Security

HTTPS and HSTS enforced, SPF and DMARC present, and the MCP flow scopes access to one quote with seven-day expiry and revocable tokens. Missing CSP, frame protection and three other headers; no SOC 2 or audit claim.

Screenshots

1 / 1

Menlo Insurance Customer MCP

About this app

Menlo Insurance Customer MCP is a hosted integration for customer-selected AI assistants and compatible applications. Its 11 deterministic tools support insurance product discovery and customer-authorized quote workflows. Menlo does not supply an AI model or autonomous agent. Public discovery is available without customer credentials; private customer access requires OAuth and customer consent. Menlo Insurance Services provides property and casualty insurance services in California and Texas from 406 Utah Street, San Francisco, CA 94110. MCP access and quote requests are free; policy premiums are separate, variable and subject to insurer underwriting. This listing does not promise coverage, a binding quote or policy issuance. Documentation and connection instructions are available at the linked website.

Key features

1MCP server at menloinsurance.com/mcp with one shared address and no client ID or secret required
2Eleven deterministic tools for product discovery and customer-authorized quote workflows
3Per-quote OAuth consent with seven-day expiry and browser-side revocation at /mcp/connect
4Documented setup for ChatGPT, Claude, Claude Code, Codex CLI, Cursor and VS Code
5Broker review for quotes started outside ChatGPT and Claude before any carrier contact
6Sensitive data (SSN, documents, signatures) stays on menloinsurance.com, never in chat

Target customer profile

Drawn from the product itself, not from a survey.

Scout Review

Written by AI from measured evidence, scored out of 100.

Menlo Insurance Customer MCP is a narrow, well-executed integration that lets ChatGPT, Claude and other MCP clients drive a real commercial insurance quote — discovery without login, quote work with scoped OAuth consent, and a broker review step for anything outside the two flagship assistants. The consent model is the most thoughtful part: one quote at a time, seven-day expiry, sensitive fields never leave menloinsurance.com. The landing page is clean and the code snippets are copy-ready. The problems are mundane and fixable: mobile LCP at 6.5s, half the security headers missing, dense text walls, no a11y statement, no demo. It's a competent product in a category where 'competent' is rarer than it should be.

Measure by measure

Usability

One shared MCP address, no client ID or secret, and copy-paste snippets for ChatGPT, Claude, Cursor, Codex and VS Code. Public discovery needs no login; quote access is a scoped OAuth consent. No demo or trial flow.

Design

Clean, confident landing page: dark-green hero, consistent navy code blocks for six MCP clients, disciplined whitespace and a scannable footer. Dense intro and privacy paragraphs and undifferentiated integration headings hold it back.

Speed

Lighthouse mobile 63/100 with LCP 6.5s and TBT 523ms; desktop 74/100. CLS is excellent at 0.002 and FCP is quick at 1.2s, but the mobile LCP is well past the 2.5s threshold.

Security

Accessibility

Lighthouse accessibility audit scores 100/100. Responsive layout, persistent phone CTA, readable contrast on code blocks. No published WCAG statement or multi-language support, so the ceiling is capped.

Growth

Real, specific wedge: MCP-native quote workflows for commercial P&C in CA and TX, with per-quote consent and broker review. Narrow segment, no network effects, but a defensible niche incumbents won't bother copying.

Improvement areas

Mobile Lighthouse performance 63/100 with LCP 6.5s — needs preloading and JS deferral
Security headers incomplete (6/12): missing CSP, frame protection, X-Content-Type-Options, Referrer-Policy, Permissions-Policy
No accessibility statement or WCAG conformance claim despite a 100/100 audit
No demo, sandbox or trial path — users must connect a real assistant to evaluate the flow
Dense unbroken paragraphs in intro and privacy sections hurt scannability

Design & Usability

The landing page does the hard part well: it explains a genuinely technical integration (MCP over HTTP) without drowning you in jargon, and the six client snippets are copy-ready with consistent styling. The hero is confident, the footer is dense but organized, and the shared server address with no client ID is a real usability win. Where it slips is density — the intro and privacy sections are unbroken paragraphs, and the six integration methods are distinguished only by their headings, so scanning is harder than it needs to be. There's also no demo, sandbox or trial path; you either connect a real assistant or you don't. For a product whose whole pitch is 'try it in the AI app you already use,' a 30-second walkthrough would earn its keep.

Speed & Security

Speed is the weakest measured dimension: Lighthouse mobile 63/100 with LCP 6.5s and TBT 523ms, though desktop lands at 74/100 and CLS is a near-perfect 0.002. The mobile LCP is the number to fix — it's more than double the 'good' threshold. Security is a mixed bag. The good: HTTPS and HSTS enforced, SPF and DMARC configured, and the MCP authorization model is unusually careful — one quote at a time, seven-day expiry, revocable tokens, sensitive fields kept on menloinsurance.com. The bad: CSP, frame protection, X-Content-Type-Options, Referrer-Policy and Permissions-Policy are all absent, giving a 6/12 header checklist. No SOC 2, ISO or pen-test claim appears on the reviewed pages. For an insurance product handling applications, the missing headers are the cheapest fix on this list.

Accessibility & Growth

Accessibility is the standout: a 100/100 Lighthouse audit, responsive layout, readable contrast on the code blocks, and a persistent phone CTA for anyone who'd rather talk to a human. The gaps are documentation, not implementation — no WCAG conformance statement and no multi-language support, which for a California and Texas insurer serving diverse small businesses is a real miss. Growth is a judgement call. The wedge is specific: MCP-native quote workflows for commercial P&C, with per-quote consent and broker review for non-ChatGPT/Claude clients. That's a genuinely differentiated position — most 'AI for insurance' plays are chat wrappers, not deterministic tools over a live quote. The limits are structural: two states, no network effects, and a category where incumbents move slowly but have distribution. It's a credible niche, not a rocket.

Conclusion

If you're a California or Texas small business already living in Claude or ChatGPT and you've been putting off a commercial quote, this is a genuinely low-friction way in — no client ID, no account creation for discovery, and you can revoke access from a browser tab. If you're outside those two states, or you want a sandbox to poke at before committing, you'll be waiting. The product is honest about its limits: nothing binds, prices are indicative, and a broker reviews anything that isn't ChatGPT or Claude. That honesty is worth more than a flashier landing page. Fix the mobile LCP, add the missing headers, and write one paragraph about accessibility, and this moves from 'interesting' to 'recommended.'

Suggestions for improvement

Cut mobile LCP from 6.5s — preload the hero, defer non-critical JS, and re-run Lighthouse until it's under 2.5s.
Add CSP, X-Content-Type-Options, Referrer-Policy, Permissions-Policy and frame protection; the checklist is 6/12 and the fixes are cheap.
Publish a short WCAG conformance statement — you're already at 100/100, so the documentation is free credibility.
Add a 30-second demo video or sandbox mode so people can see the quote flow before connecting a real assistant.

How we score

Alternatives comparison

Named competitors, point by point. Nobody paid to appear here or to be left out.

	Menlo Insurance Customer MCP
Integration protocol	MCP over HTTP (Model Context Protocol)	Not specified in provided details
Authentication model	OAuth with per-quote consent, 7-day expiry, revocable	Not specified in provided details
Public discovery without login	Yes — product and state discovery available without credentials	Not specified in provided details
Supported AI clients	ChatGPT, Claude, Claude Code, Codex CLI, Cursor, VS Code, Windsurf	Not specified in provided details
Geographic coverage	California and Texas	Not specified in provided details
Cost to use the integration	Free; policy premiums separate and subject to underwriting	Not specified in provided details