mocktomer

Developers · MCP server

Your AI assistant can now hire our shoppers

mocktomer ships a remote MCP server. Tell Claude, ChatGPT, or Cursor to test your site — human personas browse it live, AI shoppers score how the assistants see it, and the full report comes back to the conversation. Works in chats, agents, and CI pipelines.

Create an API keyFree on every plan — runs spend your normal credits.

Two-minute setup

Paste one config, then just ask

Claude Code

claude mcp add --transport http mocktomer \
  https://mocktomer.ai/mcp \
  --header "Authorization: Bearer mk_live_YOUR_KEY"

Cursor · any JSON config

{
  "mcpServers": {
    "mocktomer": {
      "url": "https://mocktomer.ai/mcp",
      "headers": { "Authorization": "Bearer mk_live_YOUR_KEY" }
    }
  }
}

ChatGPT — developer mode

Settings → Connectors → Advanced → Developer mode
→ Add connector
   URL:  https://mocktomer.ai/mcp
   Auth: Bearer mk_live_YOUR_KEY

Then: “Test mystore.com with 5 shoppers and tell me why people bail at checkout.” The assistant adds the site, launches, polls, and reads the report back — in one conversation.

The tool surface

Eight tools — the whole loop

list_sitessites in the workspace, with status + persona counts
add_sitecrawl a new site, understand the business, generate the persona cast
launch_testsend human + AI shoppers in; returns run_id instantly (~10 min runs)
get_run_statuslive per-shopper progress + AI pass state
get_reporthealth score, summary, every finding with fixes and real quotes
get_ai_readiness0–100 score + which competitors ChatGPT/Claude/Gemini recommend instead
triage_findingmark findings fixed / ignored / out of scope
rerun_verifyfocused follow-up run that verifies your fixes

Security model

  • • Keys hashed at rest, shown once, revocable within a minute
  • • Optional auto-expiry (30 / 90 / 180 days)
  • • Per-key daily run cap — a circuit breaker for leaked keys and looping agents
  • • Keys never carry admin powers, even on admin accounts
  • • Every launch spends normal credits with the dashboard's exact guardrails
  • • Shoppers never place real orders — same hard rule as everywhere

FAQ

Quick answers

What can an AI assistant actually do through the mocktomer MCP server?

Everything in the core loop: list and add sites, launch tests with human personas and AI shoppers, poll run status, read the full friction report (findings, fixes, shopper quotes), read the AI Readiness Score with the competitors assistants recommend instead, triage findings, and launch verification re-runs.

How is it secured?

Scoped API keys minted in Settings: hashed at rest, shown once, optionally auto-expiring, instantly revocable, with a per-key daily run cap as a circuit breaker against leaked keys and runaway agent loops. Keys never carry admin powers, and every launch spends the account's normal credits with the same guardrails as the dashboard.

Does it work with ChatGPT and Cursor too, or just Claude?

Any MCP client that supports remote servers over Streamable HTTP with a bearer header: Claude Code, ChatGPT developer mode, Cursor, and the Claude Messages-API MCP connector all work today.

How much does it cost?

The MCP server is free on every plan — runs launched through it spend your normal human and AI shopper credits, exactly like runs launched from the dashboard. Free accounts get 10 human shoppers monthly.

Create your key