Cycling through all six. Tap any point to stop.

Measured on six things

Give AI your phone number. With the number you get from callmemcp.com we basically support ANY AI agent, from chatgpt dots to claude to grok bot.

Security

HTTPS is enforced but the header probe found no HSTS, CSP, frame protection, Referrer-Policy or Permissions-Policy, and no DMARC — 3/12 checklist points. Privacy posture is stated clearly (ElevenLabs transcription, no audio stored), but hardening is thin.

Screenshots

1 / 3

CallMeMCP

About this app

Give AI your phone number. With the number you get from callmemcp.com we basically support ANY AI agent, from chatgpt dots to claude to grok bot. When your agent is blocked or done, it rings your phone with a real call, reads its question aloud and sends your spoken answer back to it as text. So Claude Code, Codex, Cursor, Gemini CLI, Grok Bot or any MCP client keeps working while you are in the car, on a walk with AirPods, or away from the desk. Agents can also text you and wait for your reply, and an answered call can stay open for a short back and forth. Setup takes a minute: tell your agent "install callmemcp.com", or add https://callmemcp.com/mcp as an MCP server. The /call-me app for iPhone or Android gives you your number. The demo line 5550001234 answers automatically, so you can try it before installing anything.

Key features

1Real phone call to your device that reads the agent's question aloud and returns your spoken answer as text.
2One-line MCP setup (https://callmemcp.com/mcp) with no account or API key required.
3Per-agent text threads, 2,000-character briefings, and a live demo line at 5550001234.

Target customer profile

Drawn from the product itself, not from a survey.

Demographic

Developers who run AI coding agents (Claude Code, Codex, Cursor) and step away from the desk

Pain points

The agent stops on a question or a permission prompt and nobody answers until you are back at the screen.

Primary needs

A real phone call that reads the question aloud, takes a spoken reply, and sets up in one line over MCP.

Scout Review

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

CallMeMCP is a focused, well-explained tool for a problem developers actually have: agents that stall the moment you leave the desk. The setup is genuinely one line, the demo line lets you test before installing, and the site loads at a perfect Lighthouse 100. The privacy write-up is candid about ElevenLabs handling your voice. Against that: the security headers are nearly absent, the app UI is plain, documentation leans on GitHub, and the store listing is filed under a different app's name — a trust problem for a product asking for your phone number.

Measure by measure

Usability

Setup is genuinely one line: add https://callmemcp.com/mcp or run the claude mcp add command. No account, no API key, and a demo line that answers so you can test before installing. Per-client docs live on GitHub, not the site.

Design

Judging the app UI from store screenshots: minimal utility styling, 4 screenshots, one language. The landing page is sharper than the app — dark developer aesthetic, clean code blocks, tidy 4-step call diagram.

Speed

Lighthouse mobile 100/100 with LCP 1.6s, CLS 0 and TBT 0ms; desktop also 100/100. A static, featherweight landing page that loads before you finish blinking.

Security

Accessibility

Lighthouse accessibility 100/100 on automated checks. Beyond that: one app language, no published accessibility statement, but native iOS/Android call screens and AirPods answering are solid platform-level wins.

Growth

A real, sharply named problem — agents stall on prompts while you're away — with a genuine wedge over Pushover: a spoken reply, not just a buzz. Early stage (domain 0.1 years, no ratings), but the MCP category has room.

Improvement areas

Security headers are largely absent (no HSTS, CSP, frame protection, DMARC) — 3/12 checklist points.
App Store listing is titled 'Keep Awake: AI Agents' rather than CallMeMCP, hurting trust and discoverability.
Per-client setup documentation lives on GitHub instead of the product site; app is localized in one language only.

Design & Usability

CallMeMCP's landing page is the strongest part of the package: a dark developer aesthetic, a clean four-step diagram showing agent → phone → voice → transcript, and copy that actually explains the loop. The app itself, judged from its four store screenshots, is plain utility styling — functional, not designed. Usability is where the product earns its keep: no account, no API key, one MCP URL or a single `claude mcp add` command, and a demo line that answers so you can hear it work before installing anything. The weak spot is documentation depth — per-client setup for Codex, Cursor and the rest is pushed to GitHub, so the site tells you it exists without showing you.

Speed & Security

Speed is the standout measured result: Lighthouse 100/100 on both mobile and desktop, LCP 1.6s, CLS 0, TBT 0ms. The site is static and lean, and nothing in the evidence suggests lag. Security is the opposite story. HTTPS is enforced, but the header probe found no HSTS, no CSP, no frame protection, no Referrer-Policy, no Permissions-Policy and no DMARC — 3 of 12 checklist points. To their credit, the privacy explanation is unusually candid: voice is transcribed live by ElevenLabs, audio is never recorded or stored, only text transcripts persist, and deleting your number wipes the data. That transparency is worth something, but it is not a substitute for transport hardening on a service that rings your phone.

Accessibility & Growth

Accessibility scores 100/100 on Lighthouse's automated audit, which covers contrast, labels and structure but not the whole of WCAG. The app ships in one language, and there is no accessibility statement on the site — normal for an early indie tool, not a scandal. Growth is the interesting question. The problem is real and specific: coding agents stop on permission prompts and questions while the developer is in the car or on a walk. The wedge against Pushover is concrete — Pushover buzzes your wrist, CallMeMCP reads the question aloud and takes a spoken answer back as text, with a 2,000-character briefing limit and per-agent threads. The category is expanding as MCP clients multiply, but the moat is thin: any notification tool could add voice, and trust in a brand-new domain is the real ceiling.

Conclusion

If you run long agent sessions and hate coming back to a stalled prompt, this is a clever, low-friction answer — install it, call the demo line, and see if the round-trip feels right. If you are handing a phone number to a service, wait for the security headers and the store listing to catch up with the pitch. The idea is stronger than the plumbing right now.

Suggestions for improvement

Add HSTS, CSP, frame protection and DMARC — a service that rings your phone should not score 3/12 on headers.
Fix the App Store listing name and metadata so it matches CallMeMCP instead of 'Keep Awake: AI Agents'.
Host per-client setup guides (Codex, Cursor, Gemini CLI) on the site instead of routing everything to GitHub.
Publish a short security/privacy page covering ElevenLabs data handling and add an accessibility statement.

How we score

Alternatives comparison

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

	CallMeMCP	Pushover
How the agent reaches you	Real phone call that reads the question aloud	Push notification to your phone
Can you reply without touching the screen	Yes — spoken answer transcribed back to the agent as text	No — notification only, reply requires opening the app
Setup effort	One MCP server URL or one-line CLI command; no account or API key	Create account, install app, generate API token, wire into scripts
Pricing model	Subscription started in-app via Apple/Google	One-time purchase per platform
Works with AI coding agents out of the box	Yes — Claude Code, Codex, Cursor, Gemini CLI, any MCP client	No native agent integration; requires custom scripting